Search

In this work, we introduce Vid2Seq, a multi-modal single-stage dense event captioning model pretrained on narrated videos which are readily-available at scale. The Vid2Seq architecture augments a language model with special time tokens, allowing it to seamlessly predict event boundaries and textual descriptions in the same output sequence. Such a unified model requires large-scale training data, which is not available in current annotated datasets. We show that it is possible to leverage unlabeled narrated videos for dense video captioning, by reformulating sentence boundaries of transcribed speech as pseudo event boundaries, and using the transcribed speech sentences as pseudo event captions. The resulting Vid2Seq model pretrained on the YT-Temporal-1B dataset improves the state of the art on a variety of dense video captioning benchmarks including YouCook2, ViTT and ActivityNet Captions. Vid2Seq also generalizes well to the tasks of video paragraph captioning and video clip captioning, and to few-shot settings. Our code is publicly available at https://antoyang.github.io/vid2seq.html., Comment: CVPR 2023 Camera-Ready; Project Webpage: https://antoyang.github.io/vid2seq.html ; 18 pages; 6 figures

International audience

Summary The study presents a comparative analysis of eight Vegetation Indices (VIs) used to examine vegetation greenness over the northern coasts of Iceland. The geographical extent of the study area is set by the coordinates of the two fjords, Eyjafjörður and Skagafjörður, notable for their agricultural significance. Vegetation in Iceland is fragile due to the harsh climate, climate change, overgrazing and volcanic activity, which increase soil erosion. The study was conducted on a Landsat TM image using SAGA GIS as a technical tool for raster bands calculations. The NDVI dataset shows a range from -0.56 to 0.24, with 0 indicating ‘no vegetation’, and negative values – ‘other surfaces’ (e.g. rocks, open terrain). The DVI, compared to the NDVI, shows statistically non-normalized values ranging from -112 to 0, with extreme negative values while the coastal vegetation areas are badly distinguished from the water areas. The NRVI shows an extent from -0.24 to 0.48 with higher values for vegetation. The NRVI reduces topographic, solar and atmospheric effects and creates a normal data distribution. RVI shows a range in a dataset from 0.2 to 3.2 with vegetation in the river valleys clearly visible and depicted, while the water areas have values 0.8 to 1.0. The CTVI shows corrected TVI, in a data range -0.10 to 1.10, as the dataset of NDVI were negative. The TVI dataset ranges from 0.44 to 0.80 with the ice-covered areas and glaciers distinguishable and water values within a range from 0.60 to 0.64 and the vegetation from 0.60 to 0.44. The TTVI dataset ranges from 0.40 to 0.80 performing similarly to the TVI, but more refined with vegetation values 0.64 to 0.68. SAVI dataset ranges from -0.80 to 0.30 with minimized effects of soil on the vegetation through a constant soil adjustment factor added into the NDVI formula. The paper presents a comparison of eight VIs for Arctic vegetation monitoring. The overall behavior of SAGA GIS in calculation and mapping of the VIs is effective in terms of their use for vegetation mapping of the region.

Video question answering (VideoQA) is a complex task that requires diverse multi-modal data for training. Manual annotation of question and answers for videos, however, is tedious and prohibits scalability. To tackle this problem, recent methods consider zero-shot settings with no manual annotation of visual question-answer. In particular, a promising approach adapts frozen autoregressive language models pretrained on Web-scale text-only data to multi-modal inputs. In contrast, we here build on frozen bidirectional language models (BiLM) and show that such an approach provides a stronger and cheaper alternative for zero-shot VideoQA. In particular, (i) we combine visual inputs with the frozen BiLM using light trainable modules, (ii) we train such modules using Web-scraped multi-modal data, and finally (iii) we perform zero-shot VideoQA inference through masked language modeling, where the masked text is the answer to a given question. Our proposed approach, FrozenBiLM, outperforms the state of the art in zero-shot VideoQA by a significant margin on a variety of datasets, including LSMDC-FiB, iVQA, MSRVTT-QA, MSVD-QA, ActivityNet-QA, TGIF-FrameQA, How2QA and TVQA. It also demonstrates competitive performance in the few-shot and fully-supervised setting. Our code and models are publicly available at https://github.com/antoyang/FrozenBiLM., NeurIPS 2022 Camera-Ready; Project Webpage: https://antoyang.github.io/frozenbilm.html; 25 pages; 5 figures

International audience; This paper examines spatial variations in the geomorphology of the Ninety East Ridge (NER), located in the Indian Ocean. The NER is an extraordinary long linear bathymetric feature with topography reflecting complex geophysical setting and geologic evolution. The research is based on a compilation of high-resolution bathymetric, geological, and gravity datasets clipped for the study area extent (65° - 107°E, 35°S - 21°N): General Bathymetric Chart of the Oceans (GEBCO), Earth Gravitational Model (EGM2008, EGM96). The submarine geomorphology of the NER was modeled by digitized cross-sectional profiles using Generic Mapping Tools (GMT). The availability of the method is explained by 1) the free datasets; 2) the open source GMT toolset; 3) the available tutorials of the GMT and the codes explained in this work. Three segments of the NER were selected, digitized, and modeled: 1) northern 89°E, 7°S to 90°E, 7°N; 2) central 88.4°E, 14.7°S to 88.8°E, 8.2°S; 3) southern 87.9°E, 17°S to 87.5°E, 27°S. Measured depths were visualized in graphs, compared, and statistically analyzed by the histograms. The northern segment has a steepness of 21.3° at the western slopes, and 14.5° at the eastern slope. The slopes on the eastern flank have dominant SE orientation. The central segment has a bell-shaped form, with the highest steepness comparing to the northern and southern segments. The eastern flank has a steepness of 49.5°. A local depression at a distance of 50 km off from the axis (90°E) continues parallel to the NER, with the shape of the narrow minor trench. The western slope has a steepness of 57.6°, decreasing to 15.6°. The southern segment has a dome-like shape form. Compared to the northern and central segments, it has a less pronounced ridge crest, with a steepness of 24.9° on the west. The eastern flank has a steepness of 36.8° until 70 km, gradually becoming steeper at 44.23°. A local minor trench structure can be seen on its eastern flank (100 km off the axis). This corresponds to the very narrow long topographic depressions stretching parallel to this segment of the NER at 90.5°E. The study contributes to regional geographic studies of Indian Ocean geomorphology and cartographic presentation of GMT functionality for marine research and oceanographic studies.

Mapping seafloor geomorphology is a complex task requiring the integration of advanced cartographic technologies and high-resolution spatial data. This paper provides a comparative geomorphological analysis of the Hellenic Trench (HT) and the Pliny Trench (PT) located in the Eastern Mediterranean Sea, southern Greece. These trenches were formed as a result of the tectonic plate subduction in the Eastern Mediterranean Sea: the northward motion of the African and Arabian plates, complicated by the regional geological settings, such as active faults and earthquakes, which resulted in their different geomorphological forms and bathymetric shapes. Technically, this paper presents an example of the advanced scripting mapping by scripting the cartographic toolset of Generic Mapping Tools (GMT), which performs mapping through shell scripts. The maps are based on the high-quality topographic, geological and geophysical data: GEBCO, EGM96, geoid, and marine free-air gravity fields. The GMT builds upon the modules used for data processing. The region was subsetted by ‘grdcut’, analysed by the Geospatial Data Abstraction Library (GDAL) (gdalinfo utility), and visualized by ‘grdimage’. Two segments of the trenches formed in a condition of varying geological and geophysical settings, have been modelled, visualized and compared, as representative cross-sections. As a result of the automated digitizing, spatial interpolation and sequential aggregating of GMT codes, the segments of the cross-sections were represented. The HT (Ionian segment) has an asymmetric one-sided shape; a steepness of 56.8° on the NE side and 7° on the SW flank. The PT has a more symmetric view; a steepness of 42.14° on its NW flank and 26.66° on its SE flank. The PT has a clear peak of the depths at -2600 to -2800 m and the most representative data range at -5000 to -6000 m. The HT has a bimodal data distribution with two peaks. The most frequent data for HT is -3200 m to -3400 m. Compared to PT, the HT is deeper. The histogram shows the peak of data for HT in the interval between -3200 to -3400 m (135 samples) while the PT shows the peak of data in the interval at -2600 to -2800 m (310 samples). Besides, 105 samples of the HT have depths exceeding 4000 m, while only 20 samples were recorded for PT in the same interval. This paper contributes to the geomorphological studies of the general Eastern Mediterranean Sea region, particularly relating to regional seafloor mapping of the Hellenic and Pliny trenches., Mapiranje geomorfologije morskoga dna složen je zadatak koji zahtijeva integraciju naprednih kartografskih tehnologija i prostornih podataka visoke rezolucije. Ovaj rad donosi usporednu geomorfološku analizu Helenskoga i Plinijskoga jarka smještenih u istočnome Sredozemlju, na jugu Grčke. Te su strukture nastale kao rezultat podvlačenja tektonske ploče, tj. kretanja afričke i arapske ploče prema sjeveru, usložnjenoga regionalnim geološkim postavkama poput aktivnih rasjeda i potresa, što je rezultiralo njihovim različitim geomorfološkim oblicima i batimetrijom. Prikazano je njihovo napredno kartiranje primjenom alata Generic Mapping Tools (GMT), koji koriste programske skripte. Karte se temelje na visokokvalitetnim topografskim, geološkim i geofizičkim podatcima, kao što su GEBCO, EGM96, geoid te gravitacijsko polje na razini mora. GMT se nadovezuje na module koji se koriste za obradu podataka. Područje je podijeljeno skriptom grdcut, a na temelju podataka iz zbirke apstraktnih geoprostornih podataka (GDAL) (skripta gdalinfo), te je na kraju vizualizirano (skripta grdimage). Dva dijela jarka oblikovana su u različitim geološkim i geofizičkim postavkama te su uspoređeni njihovi reprezentativni presjeci. Kao rezultat automatizirane digitalizacije, prostorne interpolacije i sekvencijalnoga agregiranja GMT kodova predstavljeni su segmenti presjeka. Helenski jarak (jonski segment) ima asimetričan jednostrani oblik, nagiba 56,8° na sjeveroistočnoj i 7° na jugozapadnoj strani. Plinijski je jarak simetričniji, nagiba 42,14° na SZ i 26,66° na JI boku. Plinijski jarak ima jasno izraženo tjeme od –2600 do –2800 m, a najreprezentativniji su podatci na –5000 do –6000 m. Helenski jarak ima bimodalnu distribuciju podataka s dvama tjemenima, a najčešće su dubine od –3200 do –3400 m. Taj je jarak dublji. Histogram prikazuje okupljanje podataka od –3200 do –3400 m, dok Plinijski (135 uzoraka) prikazuje maksimalne vrijednosti od –2600 do –2800 m (310 uzoraka). Osim toga, 105 uzoraka Helenskoga jarka ima dubine veće od –4000 m, dok je u istome intervalu za Plinijski zabilježeno samo 20 uzoraka. Članak doprinosi geomorfološkim studijama regije istočnoga Sredozemnog mora općenito i regionalnomu mapiranju morskoga dna, posebno Helenskoga i Plinijskoga jarka.

Vegetation of Cameroon includes a variety of landscape types with high biodiversity. Ecological monitoring of Yaoundé requires visualization of vegetation types in context of climate change. Vegetation Indices (VIs) derived from Sentinel-2 multispectral satellite image were analyzed in SAGA GIS to separate wetland biomes, as well as savannah and tropical rainforests. The methodology includes computing 6 VIs: NDVI, DVI, SAVI, RVI, TTVI, CTVI. The VIs shown correlation of data with vegetation distribution rising from wetlands, grassland, savanna, and shrub land towards tropical rainforests, increasing values along with canopy greenness, while also being inversely proportional to soils, urban spaces and Sanaga River. The study contributed to the environmental studies of Cameroon and demonstration of the satellite image processing.

The study presents an analysis of the sediment thickness compared with bathymetric and geophysical settings in the Bay of Bengal and Andaman Sea, Indian Ocean. It uses a combination of the high-resolution data: topographic GEBCO, satellite and marine gravity anomalies, EGM2008 geoid and GlobSed to visualize the correlation between relief, gravity and trends in continent-ocean sediment transport. The results include thematic maps and 3D model showing increased sediment thickness in the Bengal Fan (8,0 to 8,2 km) in NE direction with maximum in Ganges Fan (16,2 km), and southward decrease in the Andaman Sea from Irrawaddy river mouth (6-7 km) to the Strait of Malacca (1-2 km). All maps and 3D model have been plotted by Generic Mapping Tools (GMT) cartographic scripting toolset version 6.0.0.

Cartographic visualisation is a central concept in geoinformatics, and Generic Mapping Tools (GMT) functionality provides a variety of modules for effective mapping. However, due to its specific scripting approach, there is not enough reported experience of GMT mapping, comparing to traditional GIS. This contribution introduces steps that can be taken to perform cartographic mapping and modelling using GMT. Geographically, this paper investigates the Izu-Bonin Trench in the Pacific Ocean. The aim was to compare its geomorphology in two segments, and each was modelled by a series of profiles. The comparative analysis shows that the southern segment is deeper and is a more precisely V-shaped form with a steeper gradient slope. The northern part has an asymmetric slope with submarine terraces to the west and a straight shape to the east. The northern profile is based on 407 samples (13.5%) at depths of −5,600 to −5,800 m, followed by 304 samples at −5,800 to −6,000 m (10%). The southern histogram has a bimodal distribution with two peaks: 523 samples (20%) with depths of −5,800 to −6,000 m. The second peak (10%) is on the Bonin Ridge. The GMT proved to be an effective instrument for marine geological mapping, 3D and 2D modelling, statistical analysis and graphical plotting.

My CIFRE thesis is a continuation of the theses conducted by Tsirizo Rabenoro (Rabenoro, 2015), Cynthia Faure (Faure, 2018), Florent Forest (Forest, 2021). The objective of this thesis is to develop a methodology to understand and highlight specific typologies of aircraft engine operation during reception tests carried out on test benches, and to help Safran Aircraft Engines' engineers in the design of their engines.From a theoretical point of view, the main challenge of this thesis is to model and explain the observed and unobserved physical phenomena using statistical methods and to interpret the cause using a (small) part of the explanatory variables. This work is therefore placed in the supervised and unsupervised framework, with a strong constraint, that of constructing interpretable models: we seek to indicate the contribution of each variable to the construction of the model, which is what we call measuring the importance of the variables. In addition, we seek to improve interpretability by constructing parsimonious models, i.e. such that the variables which do not contribute to their construction are excluded. In the unsupervised framework, methods for partitioning groups of observations, or \textit{clustering} methods, will be studied. The solution we are looking for must be able to be used in high dimension, must be interpretable and must be able to take into account the group structure of the variables. This type of method is known as sparse clustering methods. In the supervised framework, we propose methods modelling a phenomenon described by continuous (respectively categorical) variables, i.e. regression (respectively classification) methods. The same constraints are imposed as in the unsupervised case (handling high dimensional data, being sparse, indicating the importance of the variables, taking into account the group structures of the variables). Moreover, let us note that we want to model a physical phenomenon with the aim of explaining its principles and that we are therefore not (directly) interested in the predictive aspects. This point is extremely important because we will see that explaining a phenomenon and explaining the prediction of the associated model are two goals that can be contradictory.A specific feature of sparse algorithms is that they depend on a parameter to be adjusted (in the same way as clustering methods). Different values of this parameter give rise to different models and it is necessary to choose between all these models. Thus, it is essential to have an efficient model selection method. It must be emphasised that model selection is a major challenge in unsupervised clustering. Indeed, there is no universally accepted method to evaluate clustering results for the obvious reason that there is no ground truth against which the results can be compared. This is also the case when one is interested in variable selection (sparse models) and variable importance, even in the supervised framework, as one never has the set of variables and the variable importance defining the underlying phenomenon under study.; Mon travail de thèse CIFRE s’inscrit dans la continuité des thèses menées par Tsirizo Rabenoro (Rabenoro,2015), Cynthia Faure (Faure, 2018), Florent Forest (Forest, 2021). L’objectif de cette thèse est de développer une méthodologie pour comprendre et mettre en évidence des typologies spécifiques du fonctionnement des moteurs d’avion lors de tests de réception effectués sur des bancs d’essai, et d’aider les ingénieurs métier de Safran Aircraft Engines dans l’analyse des résultats.Du point de vue théorique, l'enjeu principal de ce travail de thèse est de modéliser et expliquer les phénomènes physiques observés et non observés à l'aide de méthodes statistiques et d'en interpréter la cause à l'aide d'une (petite) partie des variables explicatives. Ce travail se place donc dans le cadre supervisé et non supervisé, avec une contrainte forte, celle de construire des modèles interprétables: on cherche à indiquer la contribution de chaque variable à la construction du modèle, c'est ce que l'on appelle mesurer l'\emph{importance des variables}. En outre, on cherche à améliorer l'interprétabilité en construisant des modèles parcimonieux ou \emph{sparses} en anglais, c'est-à-dire tels que les variables qui ne contribuent pas à leur construction en soient exclues. Dans le cadre non supervisé, des méthodes de partitionnement de groupes d'observations, ou méthodes de \textit{clustering} en anglais, vont être étudiées. La solution que nous cherchons doit pouvoir s'utiliser en grande dimension, doit être interprétable et elle doit pouvoir tenir compte de la structure de groupes des variables. Ce type de méthode est connu sous le nom de méthodes de clustering sparse. Dans le cadre supervisé, nous proposons des méthodes modélisant un phénomène décrit par des variables continues (respectivement catégorielles), c'est-à-dire des méthodes de régression (respectivement de classification). On impose les mêmes contraintes que dans le cas non supervisé (gérer des données de grande dimension, être sparse, indiquer l'importance des variables, tenir en compte des structures de groupes de variables). Par ailleurs, remarquons que nous voulons modéliser un phénomène physique dans le but d'en expliquer les principes et donc que nous ne sommes pas (directement) intéressés par les aspects prédictifs. Ce point est extrêmement important car nous verrons qu'expliquer un phénomène et expliquer la prédiction du modèle associé sont deux buts qui peuvent être contradictoires.Une caractéristique propre aux algorithmes sparses est qu'ils dépendent d'un paramètre à ajuster (au même titre que les méthodes de clustering). Différentes valeurs de ce paramètre donnent naissance à différents modèles et il est nécessaire de choisir entre tous ces modèles. Ainsi, il est primordial de disposer d'une méthode de sélection de modèle efficace. Il faut insister sur le fait que la sélection de modèle est un défi majeur en clustering non supervisé. En effet, il n'existe pas de méthode universellement admise pour évaluer les résultats du clustering pour la raison évidente qu'il n'y a pas de vérité de terrain par rapport à laquelle les résultats pourraient être comparés. C'est aussi le cas lorsque l'on s'intéresse à la sélection de variables (modèles sparses) et à l'importance de variables, même dans le cadre supervisé, car on ne dispose jamais du \emph{vrai} ensemble de variables et des \emph{vraies} importances de variables définissant le phénomène sous-jacent étudié.

Personality computing and affective computing have gained recent interest in many research areas. The datasets for the task generally have multiple modalities like video, audio, language and bio-signals. In this paper, we propose a flexible model for the task which exploits all available data. The task involves complex relations and to avoid using a large model for video processing specifically, we propose the use of behaviour encoding which boosts performance with minimal change to the model. Cross-attention using transformers has become popular in recent times and is utilised for fusion of different modalities. Since long term relations may exist, breaking the input into chunks is not desirable, thus the proposed model processes the entire input together. Our experiments show the importance of each of the above contributions, Comment: Preprint. Final paper accepted at the 17th International Conference on Computer Vision Theory and Applications (VISAPP), virtual, February, 2022. 8 pages

Human behavior understanding requires looking at minute details in the large context of a scene containing multiple input modalities. It is necessary as it allows the design of more human-like machines. While transformer approaches have shown great improvements, they face multiple challenges such as lack of data or background noise. To tackle these, we introduce the Forced Attention (FAt) Transformer which utilize forced attention with a modified backbone for input encoding and a use of additional inputs. In addition to improving the performance on different tasks and inputs, the modification requires less time and memory resources. We provide a model for a generalised feature extraction for tasks concerning social signals and behavior analysis. Our focus is on understanding behavior in videos where people are interacting with each other or talking into the camera which simulates the first person point of view in social interaction. FAt Transformers are applied to two downstream tasks: personality recognition and body language recognition. We achieve state-of-the-art results for Udiva v0.5, First Impressions v2 and MPII Group Interaction datasets. We further provide an extensive ablation study of the proposed architecture., Comment: Preprint. Full paper accepted at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Waikoloa, USA, Jan 2023. 11 pages

International audience; U ovom istraživanju razvijen je integrisani okvir za analizu podataka o Zemlji (“Big Earth data”) u kon- tekstu geomorfologije Jordana. Istraživanje se bavi korelacijom između nekoliko tematskih skupova podataka, uključujući mašinsko učenje i multidisciplinarne geoprostorne podatke. GIS kartiranje se široko koristi u geološkom mapiranju kao najadekvatnije tehničko sredstvo za vizualizaciju i anal- izu podataka. GIS aplikacije podstiču geološko prospektivno modeliranje vizualizacijom podataka usmjerenih na prognozu mineralnih resursa. Međutim, automatizacija pomoću mašinskog učenja za obradu velikih podataka o Zemlji pruža brzinu i tačnu obradu masivnih skupova podataka iz više izvora. To je moguće primjenom skriptiranja i programiranja u kartografskim tehnikama. Ova stu- dija predstavlja kombinovane metode mašinskog učenja kartografske analize i modeliranja velikih podataka o Zemlji. Cilj studije je da se analizira povezanost faktora koji utiču na geomorfološki oblik Jordana u odnosu na rasjed Mrtvog mora i geološku evoluciju. Tehnička metodologija uključuje tri nezavisna alata: 1) Generic Mapping Tools (GMT); 2) odabrane biblioteke programskog jezika R; 3) QGIS. Konkretno, GMT skriptni program korišćen je za topografsko, seizmičko i geofizičko mapiranje; QGIS - za geološko mapiranje; Programski jezik R - za geomorfometrijsko modeliranje. Shodno tome, tok rada je logično strukturisan kroz ova tri tehnička alata, predstavljajući različite kartografske pri- stupe za obradu podataka. Podaci i materijali uključuju skupove podataka iz više izvora različite rezo- lucije, prostornog opsega, porijekla i formata. Rezultati su predstavili kartografske rasporede kvalita- tivnih i kvantitativnih karata sa statističkim rezimeom (histogrami). Novost ovog pristupa objašnjava se potrebom da se smanji tehnički jaz između tradicionalnog GIS-a i skripti za kartiranje, koje je bit- no za mapiranje velikih podataka, gdje su presudni faktori brzina i preciznost rukovanja podacima i efikasna vizuelizacija postignuta mašinskom grafikom. U radu se analiziraju osnovni geološki procesi koji utiču na formiranje geomorfoloških oblika terena u Jordanu sa 3D vizualizacijom izabranog frag- menta zone rasjeda Mrtvog mora. Istraživanje predstavlja prošireni opis metodologije, uključujući objašnjenja isječaka koda iz modula GMT i primjere upotrebe R-biblioteka ‚raster‘ i ‚tmap‘. Rezultati su otkrili snažnu korelaciju između geoloških i geofizičkih karakteristika terena i geomorfoloških obrazaca. Integrisano proučavanje geomorfologije Jordana zasnovano je na skupovima podataka koji obrađuju više scenarija. Temeljnom analizom predstavljene su regionalne korelacije između geomorfološkog, geološkog i tektonskog okruženja u Jordanu. Rad je doprinio razvoju kartografskog inženjerstva uvođenjem skriptnih tehnika i regionalnim studijama Jordana, uključujući Mrtvo more kao posebnu regiju Jordana. Rezultati uključuju 12 novih tematskih mapa, uključujući 3D model.; In this research, an integrated framework on the big Earth data analysis has been developed in the context of the geomorphology of Jordan. The research explores the correlation between several thematic datasets, including machine learning and multidisciplinary geospatial data. GIS mapping is widely used in geological mapping as the most adequate technical tool for data visualization and analysis. GIS applications encourage geological prospective modeling by visualizing data aimed at the prognosis of mineral resources. However, automatization using machine learning for big Earth data processing provides the speed and accurate processing of multisource massive datasets. This is enabled by the application of scripting and programming in cartographic techniques. This study presents the combined machine learning methods of cartographic analysis and big Earth data modeling. The objective is to analyze a correlation between the factors affecting the geomorphological shape of Jordan with respects to the Dead Sea Fault and geological evolution. The technical methodology includes the following three independent tools: 1) Generic Mapping Tools (GMT); 2) Selected libraries of R programming language; 3) QGIS. Specifically, the GMT scripting program was used for topographic, seismic and geophysical mapping, while QGIS was used for geologic mapping and R language for geomorphometric modeling. Accordingly, the workflow is logically structured through these three technical tools, representing different cartographic approaches for data processing. Data and materials include multisource datasets of the various resolution, spatial extent, origin and formats. The results presented cartographic layouts of qualitative and quantitative maps with statistical summaries (histograms). The novelty of this approach is explained by the need to close a technical gap between the traditional GIS and scripting mapping, which is wider for big data mapping and where the crucial factors are speed and precision of data handling, as well as effective visualization achieved by the machine graphics. The paper analyzes the underlying geologic processes affecting the formation of geomorphological landforms in Jordan with a 3D visualization of the selected fragment of the Dead Sea Fault zone. The research presents an extended description in methodology, including the explanations of code snippets from the GMT modules and examples of the use of R libraries 'raster' and 'tmap'. The results revealed strong correlation between the geological and geophysical settings which affect geomorphological patterns. Integrated study of the geomorphology of Jordan was based on multisource datasets processed by scripting. A thorough analysis presented regional correlations between the geomorphological, geological and tectonic settings in Jordan. The paper contributed both to the development of cartographic engineering by introducing scripting techniques and to the regional studies of Jordan including the Dead Sea Fault as a special region of Jordan. The results include 12 new thematic maps including a 3D model.

International audience; The paper presents the cartographic processing of the Landsat TM image by the two unsupervised classification methods of SAGA GIS: ISODATA and K-means clustering. The approaches were tested and compared for land cover type mapping. Vegetation areas were detected and separated from other land cover types in the study area of southwestern Iceland. The number of clusters was set to ten classes. The processing of the satellite image by SAGA GIS was achieved using Imagery Classification tools in the Geoprocessing menu of SAGA GIS. Unsupervised classification performed effectively in the unlabeled pixels for the land cover types using machine learning in GIS. Following an iterative approach of clustering, the pixels were grouped in each step of the algorithm and the clusters were reassigned as centroids. The paper contributes to the technical development of the application of machine learning in cartography by demonstrating the effectiveness of SAGA GIS in remote sensing data processing applied for vegetation and environmental mapping.

International audience; This paper presents the GRASS GIS-based thematic mapping of Antarctica using scripting approach and associated datasets on topography and geophysics. The state-ofthe art in cartographic development points at two important aspects. The first one comprises shell scripting promoted repeatability of the GIS technique, increased automatization in cartographic workflow, and compatibility of GRASS with Python, PROJ and GDAL libraries which enables advanced geospatial data processing: converting formats, re-projecting and spatial analysis. The second aspect is that data visualization greatly influences geologic research through improving the interpretation between the Antarctic glaciation and surface. This includes the machine learning algorithms of image classification enabling to distinguish between glacier and non-glacier surfaces through automatically partitioning data and analysis of various types of surfaces. Presented detailed maps of Antarctic include visualized datasets from the ETOPO1, GlobSed, EGM96 and Bedmap2 projects. The grids include bed and surface elevation, ETOPO1-based bathymetry and topography, bed, ice and sediment thickness, grounded bed uncertainty, subglacial bed elevation, geoid undulations, ice mask grounded and shelves. Data show the distribution of the present-day glacier, geophysical fields and topographic landforms for analysis of processes and correlations between the geophysical and geological phenomena. Advances in scripting cartography are significant contributions to the geological and glaciological research. Processing high-resolution datasets of Southern Ocean retrieved by remote sensing methods present new steps in automatization of the digital mapping, as presented in this research, and promotes comprehensive monitoring of geological, permafrost and glacial processes in Antarctica. All maps have been plotted using GRASS GIS version 7.8. with technical details of scripts described and interpreted.

With the massive use of digital photography, large databases of images have been built during the last decades, both for industry and medicine. However, the images were generally captured with different lighting conditions, poses of the camera and types of camera. This causes difficulties when comparing images of the same scene. I have proposed methods to analyse these large databases which address these issues. They have been applied to the field of tyres for the control of their visual aspect and the study of their performances as well as to the field of healthcare for the diagnosis and the follow-up of diabetic retinopathy. These methods fall into the framework of Mathematical Morphology and Logarithmic Image processing (LIP). In particular, thanks to partnerships with academia, I have developed approaches of morphological segmentation and classification by machine learning. They have been used for the analysis of images of tyre surfaces (which can be textured or in three dimensions) in order to look for defaults. They have also served for the automatic analysis of eye fundus images from patients with diabetes. Due to the large amount of data to analyse, I have speeded up algorithms with regards to their complexity as well as their efficient programming. The program can be run in parallel over several processor cores or written in a vector way (i.e., several numbers processed at the same time in place of a single one). Thanks to the LIP model, I have proposed methods to improve colour contrast in retinal images. I have studied the properties of these functional metrics which have been defined either with the LIP-additive law, or with the LIP-multiplicative law. The first metric, which is LIP-additive, is robust to lighting variations caused by a change in the exposure time of the camera or in the source intensity. The second one, which is LIP-multiplicative, is robust to lighting variations due to changes of opacity of the captured object. These metrics are useful for pattern matching, thanks to distance maps between a reference function and an image. I have established the link between these maps of Asplund distances and Mathematical Morphology. Then, I have created the new framework of Logarithmic Mathematical Morphology which is based on the fundamental operations of erosion and dilation with the LIP-additive law. This gives them the interesting property of being adaptive to lighting variations caused by a change of camera exposure-time. Other operators, robust to these variations, have been defined in this novel framework. The latter and the LIP model have allowed the introduction of segmentation methods robust to lighting variations. Percolation techniques have also been studied with logarithmic colour contrasts. In addition, I have conceived and made a three-dimensional and multimodal acquisition system by stereovision, in the domain of the visible light and in the one of X-rays for tyre imaging. The complete chain has been studied and validated, including stereoscopic acquisition, calibration of the acquisition system and 3D reconstruction. In addition, in order to compare images of radial cuts of tyres with their 2D plan, a deformable registration method based on cubic B-splines has been successfully tested. For the longitudinal (i.e., temporal) analysis of eye fundus images, I have introduced a superimposition model made of an affine homography (i.e., a rotation, a translation and an anisotropic scaling) and one or two corrections of radial distortions depending on the number of cameras used for the acquisition. Compared to other state-of-the-art methods, in series of image pairs from public health databases captured with an interval of a year, the proposed approach gives better results. These research works form my contribution to the analysis of large image databases in industry or in medicine., Avec l’utilisation de plus en plus massive de la photographie numérique, de grandes banques de données d’images ont été constituées durant ces dernières décennies tant dans l’industrie que pour la médecine. Cependant, les images ont généralement été acquises avec différentes conditions d’éclairement, poses de l’appareil photographique et types d’appareil. Ceci cause des difficultés de comparaison entre les images d’une même scène. J’ai proposé des méthodes pour l’analyse de ces grandes banques de données qui répondent à cette problématique. Elles ont été appliquées dans le domaine des pneumatiques pour le contrôle de leur aspect visuel et l’étude de leur performancesainsi que dans celui de la santé pour le diagnostic et le suivi de la rétinopathie diabétique. Ces méthodes appartiennent aux cadres de travail de la Morphologie Mathématique et du Logarithmic Image Processing (LIP). Grâce à des collaborations académiques, j’ai développé des approches de segmentation morphologiques et de classification par apprentissage statistique. Elles ont été utilisées pour l’analyse d’images texturées et tridimensionnelles de la surface des pneumatiques, à des fins de recherche de défauts. Elles ont aussi servi à l’analyse automatique des images du fond d’œil chez les patients atteints de diabète. Du fait de la grande quantité de données à analyser, j’ai amélioré des algorithmes tant au niveau de leur complexité qu’au niveau de leur programmation efficace. Le programme peut être exécuté en parallèle sur plusieurs cœurs de processeurs et écritde manière vectorielle (i.e. plusieurs nombres traités à la fois au lieu d’un seul). Grâce au modèle LIP, j’ai proposé des méthodes d’amélioration des contrastes couleur des images rétiniennes. J’ai étudié les propriétés des métriques fonctionnelles d’Asplund définies soit avec la loi LIP-additive, soit avec la loi LIP-multiplicative. La première métrique, qui est LIP-additive, est robuste aux variations d’éclairement causées par un changement du temps d’exposition de la caméra ou de l’intensité de la source. La seconde, qui est LIP-multiplicative, est robuste aux variations d’éclairement dues à un changement d’opacité de l’objet capturé. Ces métriques sont utiles à la reconnaissance des formes, grâce aux cartes de distances entre une fonction de référence et une image.J’ai établi le lien entre ces cartes de distances d’Asplund et la Morphologie Mathématique. Puis, j’ai créé le nouveau cadre de la Morphologie Mathématique Logarithmique qui consiste à définir les opérations fondamentales d’érosion et de dilatation avec la loi LIP-additive. Ceci leur donne la propriété intéressante d’être adaptatives aux variations d’éclairement causées par un changementdu temps d’exposition de la caméra. D’autres opérateurs, robustes à ces variations, ont été définis dans ce nouveau cadre de travail. Celui-ci et le modèle LIP ont permis l’introduction de méthodes de segmentation robustes aux variations d’éclairement. Des techniques de percolations ont aussi été étudiées avec des contrastes logarithmiques couleur. En outre, j’ai conçu et réalisé un système d’acquisitions tridimensionnelles et multimodales par stéréovision dans le domaine de la lumière visible et dans celui des rayons X pour l’imagerie des pneumatiques. La chaîne complète comprenant l’acquisition stéréoscopique, le calibrage des systèmes d’acquisition et la reconstruction 3D a été étudiée et validée. De plus, afin de comparer les images de coupes radiales de pneumatiques avec leurs épures (i.e. leurs plans 2D), une méthode de recalage déformable à partir de B-splines cubiques a été testée avec succès. Pour l’analyse longitudinale (i.e. temporelle) des images du fond d’œil, j’ai introduit un nouveau modèle de superposition composé d’une homographie affine (i.e. une rotation, une translation et une mise à l’échelle anisotrope) et d’une ou deux corrections de distorsions radiales selon le nombre de caméras employées lors de l’acquisition. Comparée à plusieurs méthodes de l’état de l’art, sur des séries de paires d’images provenant de banques dedonnées de santé publique acquises avec une année d’intervalle, l’approche proposée donne de meilleurs résultats. Ces travaux de recherche constituent ma contribution à l’analyse des grandes banques d’images dans l’industrie ou en médecine.

International audience; Scripting cartographic technique is a new method of geospatial data visualization – especially with thematic mapping such as geomorphological models. The purpose of this study was to explore the use of Generic Mapping Tools (GMT) and R for geomorphological and topographic mapping of Serbia using free open datasets (DEM, SRTM/ GEBCO, OpenStreetMap). Current trend in education and research of distance-based and online-based education suggests that application of free high-resolution data for modelling and mapping by open source cartographic toolsets are more likely to result in deep geospa- tial analysis of the geomorphology of Balkans with associated geographic phenomena: hydrology, soils, vegetation, geology. Presented fragments of scripts aim to demonstrate the technical usage of R and GMT coding in cartographic workflow with a case study on Serbia. Using qualitative descriptive cartographic approach to visualize the slope, aspect and terrain elevations over the country and the ‘raster’ package provided by R, it was found that the geomorphology of Serbia spatially differs in southern and northern parts of the country re- sulting in regional geologic evolution and tectonic dynamics of the Balkan formation. The presented maps portrayed general distribution of the landforms in Serbian region of Balkans. The research contributes to the methodological development and testing of the cartographic techniques as well as geomorphological and environmental studies of Serbia.; Скриптна картографска техника је нов метод визуелизације гео- спацијалних података – посебно са тематским мапирањем као што су геоморфо- лошки модели. Циљ овог проучавања је истраживање употребе Генеричког алата за мапирање (енгл. GMT) и R за геоморфолошко и топографско мапирање Србије користећи слободне отворене скупове података (DEM, SRTM/GEBCO, OpenStreet- Map). Актуелан тренд у образовању и истраживање наставе на даљину и онлајн наставе указује на то да примена слободних података високе резолуције за моде- ловање и мапирање путем картографског алата из отворених извора, највероват- није резултира дубоком геоспацијалном анализом геоморфологије Балкана са повезаним географским феноменима: хидрологија, земљишта, вегетација, гео- логија. Приказани фрагменти скрипта имају за циљ да демонстрирају техничку употребу R и GMT кодирања у картографском процесу рада где је Србија у фоку- су истраживања случаја. Користећи квалитативни описни картографски приступ да би се визуелизовали нагиб, положај и издизање земљишта широм земље и „растер” пакет обезбеђен путем R, откривено је да се геоморфологија Србије про- сторно разликује у јужним и северним деловима земље што резултира регионал- ном геолошком еволуцијом и тектонском динамиком балканске формације. При- казане мапе показале су генералну дистрибуцију рељефа у делу Балкана у којем је Србија. Истраживање доприноси методолошком развоју и тестирању карто- графских техника као и геоморфолошким изучавањима и изучавањима животне средине у Србији.

International audience; Detailed mapping based on the high-resolution grids, such as GEBCO, ETOPO1, GlobSed, EGM-2008 is crucial for various domains of Earth sciences: geophysics, glaciology, Quaternary, sedimentology, geology, environmental science, geomorphology, etc. The study presented a GMTbased scripting technique of the cartographic data processing aimed at the comparative analysis of the bathymetry, sediment thickness, geologic objects and geophysical settings in the study area based on various datasets. The study area is located in the Ross Sea, Antarctic. The highest values of the sediment thickness over 7,500 m are dominating in the southwest segment of the Ross Sea closer to the Victoria Land, followed by the region over the Ross Ice Shelf with values between 5,500 to 7,000 m (170°-175°W). The increased sediment thickness (2,500 to 3,000 m) was also mapped seen in the region NE off the Sulzberger Bay (70-75°S to 140-155°W), caused by the closeness of the Marie Bird Land ice coasts. A remarkable correlation between the gravity and the topography of the sea-land border in the Marie Bird Land area is well reflected in the coastal line and a set of the higher values in the free-air gravity. On the contrary, negative values (-60 to-80 mGal) are notable along the submarine troughs stretching parallel in the western part of the basin: e.g. the trough stretching in NW-SE direction in the 170°W-175°E, 65°S-68°S, between the 167°W-175°W, 70°S-72°S. Such correlations are clearly visible on the map, indicating geological lineaments and bathymetric depressions correlating with gravity grids. The paper contributes to the regional studies of the Ross Sea, the Antarctic and Polar region, and development of the technical cartographic methodologies by presenting an application of the GMT for thematic mapping.; O mapeamento detalhado com base em grades de alta resolução, como GEBCO, ETOPO1, GlobSed, EGM-2008 é fundamental para vários campos das ciências da terra: geofísica, glaciologia, quaternário, sedimentologia, geologia, ciências ambientais, geomorfologia, etc. O estudo apresenta métodos de processamento de dados de mapas com script baseado em GMT visando a análise comparativa de batimetria, espessura de sedimentos, características geológicas e condições geofísicas na área de estudo com base em vários conjuntos de dados. A Área de Estudo está localizada no Mar de Ross, na Antártica. Os valores mais altos de espessura de sedimento de mais de 7.500 m prevalecem no sudoeste do Mar de Ross próximo a Victoria Land, seguido pela área acima da plataforma de gelo de Ross em valores entre 5.500 e 7.000 m (170 ° -175 ° W). ... O aumento da espessura do sedimento (de 2.500 a 3.000 m) também foi mapeado na área nordeste da Baía Sulzberger (de 70-75 ° S a 140-155 ° W), devido à proximidade das costas de gelo da Terra Marie Bird. A notável correlação entre a gravidade e a topografia da fronteira marítima perto da Terra Marie Bird é bem refletida na linha costeira e em muitos dos maiores valores de gravidade ao ar livre. Em contraste, os valores negativos (-60 a -80 mGal) são perceptíveis ao longo das muralhas submarinas que se estendem paralelamente na parte oeste da bacia: por exemplo, uma trincheira que se estende na direção noroeste-sudeste a 170 ° W-175 ° E, 65 ° S-68 ° S, entre 167 ° W-175 ° W, 70 ° S-72 ° S Essas correlações são claramente visíveis no mapa, indicando lineamentos geológicos e depressões batimétricas correlacionadas com grades gravimétricas. O documento contribui para estudos regionais do Mar de Ross, Antártica e região polar, bem como para o desenvolvimento de metodologias técnicas de mapeamento, introduzindo a aplicação GMT para mapeamento temático.

International audience; Detailed mapping based on the high-resolution grids, such as GEBCO, ETOPO1, GlobSed, EGM-2008 is crucial for various domains of Earth sciences: geophysics, glaciology, Quaternary, sedimentology, geology, environmental science, geomorphology, etc. The study presented a GMTbased scripting technique of the cartographic data processing aimed at the comparative analysis of the bathymetry, sediment thickness, geologic objects and geophysical settings in the study area based on various datasets. The study area is located in the Ross Sea, Antarctic. The highest values of the sediment thickness over 7,500 m are dominating in the southwest segment of the Ross Sea closer to the Victoria Land, followed by the region over the Ross Ice Shelf with values between 5,500 to 7,000 m (170°-175°W). The increased sediment thickness (2,500 to 3,000 m) was also mapped seen in the region NE off the Sulzberger Bay (70-75°S to 140-155°W), caused by the closeness of the Marie Bird Land ice coasts. A remarkable correlation between the gravity and the topography of the sea-land border in the Marie Bird Land area is well reflected in the coastal line and a set of the higher values in the free-air gravity. On the contrary, negative values (-60 to-80 mGal) are notable along the submarine troughs stretching parallel in the western part of the basin: e.g. the trough stretching in NW-SE direction in the 170°W-175°E, 65°S-68°S, between the 167°W-175°W, 70°S-72°S. Such correlations are clearly visible on the map, indicating geological lineaments and bathymetric depressions correlating with gravity grids. The paper contributes to the regional studies of the Ross Sea, the Antarctic and Polar region, and development of the technical cartographic methodologies by presenting an application of the GMT for thematic mapping.; O mapeamento detalhado com base em grades de alta resolução, como GEBCO, ETOPO1, GlobSed, EGM-2008 é fundamental para vários campos das ciências da terra: geofísica, glaciologia, quaternário, sedimentologia, geologia, ciências ambientais, geomorfologia, etc. O estudo apresenta métodos de processamento de dados de mapas com script baseado em GMT visando a análise comparativa de batimetria, espessura de sedimentos, características geológicas e condições geofísicas na área de estudo com base em vários conjuntos de dados. A Área de Estudo está localizada no Mar de Ross, na Antártica. Os valores mais altos de espessura de sedimento de mais de 7.500 m prevalecem no sudoeste do Mar de Ross próximo a Victoria Land, seguido pela área acima da plataforma de gelo de Ross em valores entre 5.500 e 7.000 m (170 ° -175 ° W). ... O aumento da espessura do sedimento (de 2.500 a 3.000 m) também foi mapeado na área nordeste da Baía Sulzberger (de 70-75 ° S a 140-155 ° W), devido à proximidade das costas de gelo da Terra Marie Bird. A notável correlação entre a gravidade e a topografia da fronteira marítima perto da Terra Marie Bird é bem refletida na linha costeira e em muitos dos maiores valores de gravidade ao ar livre. Em contraste, os valores negativos (-60 a -80 mGal) são perceptíveis ao longo das muralhas submarinas que se estendem paralelamente na parte oeste da bacia: por exemplo, uma trincheira que se estende na direção noroeste-sudeste a 170 ° W-175 ° E, 65 ° S-68 ° S, entre 167 ° W-175 ° W, 70 ° S-72 ° S Essas correlações são claramente visíveis no mapa, indicando lineamentos geológicos e depressões batimétricas correlacionadas com grades gravimétricas. O documento contribui para estudos regionais do Mar de Ross, Antártica e região polar, bem como para o desenvolvimento de metodologias técnicas de mapeamento, introduzindo a aplicação GMT para mapeamento temático.

International audience; The paper discusses the geomorphology of the Sunda Trench, an oceanic trench located in the eastern Indian Ocean along the Sumatra and Java Islands of the Indonesian archipelago. It analysis difference in depths and variation in slope steepness between the two segments of the trench: southern Java transect (108.8°E 10.10°S - 113.0°E 10.75°S) and northern Sumatra transect (97.5°E 1.1°S - 101.0°E 5.5°S). The maps and geomorphological modelling were plotted using Generic Mapping Tools (GMT). The data include high-resolution grids on topography, geology, geodesy and geophysics: GEBCO, EGM2008 EGM-2008, GlobSed. The results include modelled segments, slope gradients, and cross-section profiles. The geological processes take place in the Indian Ocean at different stages of its evolution and influence the nature of the submarine geomorphology and geomorphology of the trench that differs in two segments. Java segment has a bell- shaped data distribution in contrast to the Sumatra with bimodal pattern. Java segment has the most repetitive depths at -2,500 to -5,200 m. Sumatra transect has two peaks: 1) a classic bell-shaped peak (-4,500 m to -5,500 m); 2) shelf area (0 to -1,750 m). The data at middle depths (-1,750 to -4,500 m) have less than 300 samples. The most frequent bathymetry for the Sumatra segment corresponds to the -4,750 m to -5,000 m. Comparing to the Sumatra segment, the Java segment is deeper. For depths > -6,000 m, there are only 138 samples for Sumatra while 547 samples for Java. Furthermore, Java segment has a more symmetrical geometric shape while Sumatra segment is asymmetric, one-sided. The Sumatra segment has a steepness of 57.86° on its eastern side (facing Sumatra Island) and a contrasting 14.58° on the western part. The Java segment has a steepness of 64.34° on its northern side (facing Java Island) and 24.95° on the southern part (facing the Indian Ocean). The paper contributes to the studies of the submarine geomorphology in Indonesia.; Резиме: У раду су детаљно анализиране геоморфолошке карактеристике Сундског рова, који се пружа дуж острва Суматра и Јава у оквиру индонежанског архипелага, у источном делу Индијског океана. Познато је да је геоморфологија океанских ровова настала као резултат субдукције тектонских плоча током сложене геолошке еволуције. Упркос покушајима да се детаљно проучи његова геодинамика, природа и структура рова још увек нису довољно јасни. Конкретно, геоморфолошке карактеристике рова, формиране кроз сложене интеракције геофизичких, геолошких и тектонских фактора, у фокусу су многих истраживања. Сврха овог рада је да истражи разлике у геоморфолошким карактеристикама два различита сегмента Сундског рова (јужни део дуж острва Јава и северни део дуж острва Суматра) како би се олакшало боље разумевање структуре морског дна у североисточном делу Индијског океана. Методолошки приступ овог рада заснован је на употреби GMT, односно универзалних картографских инструмената, који су примењени у циљу конструисања графичких прилога. У овом квалитативном и квантитативном истраживању израђено је осам карата и обављен је преглед бројне литературе о истраживаном подручју.Јужни, јавански сегмент рова простире се дуж линије са следећим координа- тама (почетна и крајња тачка): 108,8°Е 10,10°S - 113,0°Е 10,75°S. Северни, суматрански сегмент лоциран је дуж линије са следећим координатама (почетна и крајња тачка) 97,5°Е 1,1°S - 101,0°Е 5,5°S. Подаци укључују растерску координатну мрежу високе резолуције о топографији, геологији, геодезији и геофизици: GEBCO, EGM2008EGM-2008, GlobSed. Представљено је и визуализовано неколико тематских карата са циљем демонстрирања просторне варијације у дистрибуцији геофизичких поља на проучаваном подручју. Карте посебно илуструју геолошке процесе који су се одвијали у источном делу Индијског океана у различитим фазама његове еволуције, а које су утицале на изглед подморске геоморфологије рова. Попречни профили су дигитализовани помоћу приступа машинског учења којег нуди GMT. Упоредна анализа оба сегмента извршена je визуелизацијом података у виду хистограма. Резултати истраживања показују да је геоморфологија рова зависна од локалних геолошких, геофизичких и тектонских карактеристика. Конкретно, јавански сегмент има дистрибуцију дубина у "облику звона", за разлику од суматранског са бимодалном дистрибуцијом. У сегменту Јаве најчешће је распрострањен појас дубина између -2.500 и -5.200 m. Сегмент Суматре има два јасно видљива врха: 1) врх у облику звона (-4.500 m до -5.500 m); 2) подручје шелфа (0 до -1.750 m). У поређењу са сегментом Суматре (северни), сегмент Јаве (јужни) је дубљи. На пример, за дубине преко -6.000 m постоји само 138 узорака за сегмент22Суматре, док постоји 547 узорака за сегмент Јаве. Даље, јавански сегмент има симетричан геометријски облик, док је сегмент Суматре углавном асиметричан. Сегмент Суматре има нагиб од 57,86° на источној страни (страна окренута према острву Суматра) и 14,58° на западној страни. Сегмент Јаве има нагиб од 64,34° на северној страни (страна окренута према острву Јава) и 24,95° у јужном делу (према Индијском океану).Значај картографске визуелизације је добро познат, јер тачно представљање података помаже да се изврши корелација и укаже на разлике у дистрибуцији геолошких и географских објеката, процеса и појава. Најчешће, традиционални GIS софтвери преовлађују у картографским истраживањима. Овај рад указује на ефикасну функционалност софтвера GMT за визуализацију геофизичких и геолошких података и за проучавање геоморфолошких карактеристика. Нејасност и неразумевање геолошких и геодинамичких процеса често проистиче из недостатка картографске визуализације која покрива одређено истраживачко подручје у правилној пројекцији и палети боја. Да би се повећала ефикасност у анализи података, представљене карте конструисане су у три различите пројекције: поликонусној, меркаторовој и азимутној пројекцији Ламберта, док су попречни профили урађени са декартовим xy координатама. Значај овог рада састоји се у мултидисциплинарном приступу који комбинује савремена картографска представ- љања података и геолошку анализу индонежанског региона. Резултати ове студије показују да су коришћење картографских метода за обраду геоинформација драгоцено средство за даља геофизичка и геоморфолошка моделовања и помоћ у бољем разумевању морског дна, јединствене скривене структуре Земљине површи- не.

International audience

International audience; The study present a case study of the Generic Mapping Tools (GMT) applied for cartographic modelling, mapping and comparative analysis of the deep-sea trenches located in southwest Pacific Ocean: the New Britain Trench (NBT) and the San Cristobal Trench (SCT). The aim was to evaluate their geomorphic variation using scripting cartographic approach of GMT. The data was processed using a sequence of the GMT modules with the main module 'grdtrack' used to visualize crosssection profiles along the trenches for their geomorphological modelling. The main grid used for topographic mapping is the SRTM DEM with 15-arc second resolution. The statistical analysis shown variability in depths of both trenches by samples in two transects. The cartographic analysis demonstrated following results. The SCT is generally deeper reaching-9,000 m, while the median for the NBT less then 7,000 m. The gradient slope of SCT is more symmetric with accurate 'V' form. In a cross-section graph, the NBT landward slope is markedly asymmetric U-shaped form and has a crescent form in the east. The NBT slope dips westwards with 35° eastward, and 41° westward, while the SCT slope has 33° oceanwards and 33,69° landwards. The difference between the geomorphology of the trenches is explained by the effects of the geotectonic evolution and actual sedimentary processes affected their formation and sculptured their structure. The marine free-air gravity anomaly illustrated density anomalies at the bathymetry in the region of NBT and SCT with range

International audience; Present study is focused on the satellite image processing by means of SAGA GIS. The objective of the study is assessment and analysis of the core urban areas and its spatial distribution in the limits of the city and suburbs. The study area includes Yaoundé and its surroundings, Republic of Cameroon. The methodology includes Object Based Image Segmentation (OBIS) approach by SAGA GIS. The paper presents a methodologically structured workflow used in SAGA GIS for segmentation of the Sentinel-2A image. The segmentation techniques includes adjusting technical parameters, performing neighborhood approach and post-processing procedures (unsupervised classification, number of clusters). The OBIS model and SAGA GIS were used as main methods and machine learning techniques for image segmentation. Data include Sentinel-2A satellite image with high resolution (10 m). The image was analyzed by two approaches of cell neighborhood analysis: Moore and Neumann. The results showed following numerical parameters of the computed area: the perimeter of 1,060,560 km and an area estimated for the Yaoundé city 191,745,000 km 2. The Neumann approach demonstrated better results for image clustering. The results presented automatically detected and separated segments of the city areas and other land cover types (savannah, forests, mountains). The spectral reflectance of various land cover types on a satellite image enables to group pixels of the image into classes using segmentation technique which has an important impact on the conceptual methodology of the urban mapping. The results of the image segmentation show the average values of the Neumann approach more correct in urban area than Moore approach. The accuracy assessment demonstrated 74.63% for the core urban area by using the Neumann method. The applicability of SAGA GIS for automated methods of image processing using machine learning algorithm of OBIS is presented and the advantages are discussed. The study demonstrated the effectiveness of the high-resolution Sentinel-2A for socioeconomic studies, exemplified by urban mapping where remote sensing data serve as reliable sources of geoinformation. The advantages of the OBIS are discussed with detailed explanation of the SAGA GIS workflow.

The paper discusses geomorphology of the Sunda Trench, an oceanic trench located in eastern Indian Ocean along the Sumatra and Java Islands of the Indonesian archipelago. In particular, it analysis the difference in depths and variation in slope steepness between the two segments of the trench: the southern Java transect (coordinates 108.8°E 10.10°S to 113.0°E 10.75°S) and the northern Sumatra transect (97.5°E 1.1°S to 101.0°E 5.5°S). The thematic maps and geomorphological modelling were plotted using Generic Mapping Tools (GMT). The materials include high-resolution data on topography, geology and geophysics: GEBCO 15 arc-minute resolution grid, EGM2008 2.5 minute Earth Gravitation Model of 2008, GlobSed global 5‐arc‐minute total sediment thickness and vector geological datasets. In addition to the GEBCO-based bathymetric data, geological, topographic and geophysical maps, the results include enlarged transects for the Java and Sumatra segments, their slope gradients and cross-section profiles, derived from the bathymetric GEBCO dataset. The geomorphology framework of the Sunda Trench is largely controlled by the subduction of the Australian plate underneath the Sunda microplate. The geological processes take place in basin of the Indian Ocean at different stages of its evolution and influence the nature of the submarine geomorphology and geometric shape of the trench. Sunda Trench is seismically active part of the Pacific Ring of Fire. A large number of the catastrophic earthquakes are recorded around the trench. The histograms shows variation in depths along the segments of the Sumatra and Java. The Java segment has a bell-shaped data distribution in contrast to the Sumatra with bimodal pattern. The Java segment has the most repetitive depths at -2,500 to -5,200 m. The Sumatra transect has two peaks: 1) a classic bell-shaped peak at depths -4,500 m to -5,500 m; 2) shelf area with a peak from 0 to -1,750 m. The data at middle depths (-1,750 to -4,500 m) have a frequency -6,000 m, there are only 138 samples for the Sumatra while 547 samples for Java. Furthermore, Java segment has more symmetrical geometric shape while Sumatra segment is asymmetric, one-sided. The Sumatra segment has a steepness of 57.86° on its eastern side (facing Sumatra Island) and a contrasting 14.58° on the western part. The Java segment has a steepness of 64.34° on its northern side (facing Java Island) and 24.95° on the southern part (facing Indian Ocean). The paper contributes to the studies of the submarine geomorphology in Indonesia.

International audience; Insights on the Indian Ocean Tectonics and Geophysics Supported by GMT. This paper presented analyzed and summarized data on geological and geophysical settings about the tectonics and geological structure of the seafloor of the Indian Ocean by thematic visualization of the topographic, geophysical and geological data. The seafloor topography of the Indian Ocean is very complex which includes underwater hills, isolated mountains, underwater canyons, abyssal and accumulative plains, trenches. Complex geological settings explain seismic activity, repetitive earthquakes, and tsunami. Understanding and prognosis of the disastrous and catastrophic geological events is strongly based on correct data analysis, modelling and visualization. An important feature of this paper is mapping multi-source high-resolution data by GMT. Data include raster grids in NetCDF and GRD formats: ETOPO1, geologic and marine free-air gravity data, EGM96, age, spreading rates, and spreading asymmetry of the ocean crust by NOAA, total sediment thickness. Data were visualized by GMT modules to compare and analyze geophysical and geological settings of the Indian Ocean. Visualization revealed correlations between high bathymetric variations of the oceanic seafloor, distribution of main geological seafloor fabric: Southwest, Southeast, Mid and Carlsberg ridges. Tectonic maps were plotted to perform comparative analysis of several variables: crust age, spreading half rates (mm/yr), asymmetries in crustal accretion on conjugate ridge flanks (%), variations in the geopotential and gravimetric models. Being the warmest of the world's ocean, Indian Ocean has specific climatic conditions (repetitive monsoons, tsunamis, cyclones and storms), complex geologic seafloor structure with triple junction and unique geographic settings. Presented paper contributed to the regional studies of the Indian Ocean.

The vegetation indices (VIs) derived from the hyperspectral reflectance of vegetation are presented in this study for monitoring live green vegetation in the northern ecosystems of Iceland, along the fjords of Eyjafjörđur and the Skagafjörđur. The comparative analysis of the following VIs was performed: the NDVI, RVI, NRVI, TVI, CTVI, TTVI and SAVI. The methodology is based on the raster calculator band in a QGIS. The dataset includes a Landsat TM scene of 2013, UTM Zone 53, WGS84 captured from the GloVis. The computed bands include the NIR and R spectral bands and their combinations according to the algorithms of each of the seven VIs. The hyperspectral reflectance and crop canopy computations were applied to generate various scales of VIs and demonstrated following data range: NDVI: -0.91 to 0.65, RVI: 0.22 to 19.65, NRVI: 0.63 to 0.90, TVI: 0 to 1.12, CTVI: -0.64 to 1.07, TTVI: 0.70 to 1.18 and SAVI: -1.36 to 0.99 (roughly to 1.00). Of these, the RVI, NRVI, TVI and TTVI are adjusted to the positive values while the NDVI, CTVI and SAVI do include the negative diapason in the dataset due to the computing algorithm. The algorithms of the seven VIs are described and visualized in form of maps based on the multispectral remote sensing Landsat TM imagery identifying vegetated areas, their health condition and distribution of green areas against the bare soils, rocks, ocean water, lakes and ice-covered glaciers. The paper contributes both to the technical presentation of the QGIS functionality for the Landsat TM data processing by a raster calculator, and to the regional geographic studies of Iceland and Arctic ecosystems.

International audience; This paper concerns the Caribbean Sea submarine geomorphology and bathymetry and especially the Puerto Rico Trench and the Cayman Trough, using thematic mapping, geomorphological modelling and statistical analysis. The technical tools include Generic Mapping Tools (GMT) cartographic scripting toolset. The data include GEBCO digital bathymetric model in grid format, geopotential model of the Earth's gravity field EGM2008, marine free-air Faye gravity anomalies from a combined GEOS3/SEASAT/GEOSAT altimeter data set, sediment thickness from the GlobSed 5-arc-minute grid model and vector layers of GMT (coastlines, river network, borders). The cross-sectioning was done by the "grdtrack" module. Differences between the form of the Puerto Rico Trench and Cayman Trough presumably result from different structure and geological evolution. The geomorphology of the segment of the Puerto Rico Trench (67.5°W and 19.90°N to 64.1°W and 19.82°N) has a gentle curvature of the slope in plane (about 13° slope steepness). The slopes are steeper in the northern part (about 32°) but higher on the continental slope. The profiles of the Puerto Rico Trench are asymmetric due to the tectonic factors. The seabed of the Cayman Trough is fl at at the segment (80.0°W and 17.7°N to 78.5°W and 19.5°N). Its profile is asymmetric: northern part is steep (about 57°), southern part is about 16°. A very large negative Faye free-air gravity anomaly (up to-380 mGal) is seen in the Puerto Rico Trench, south of Cuba as well as in the northeastern part of the Cayman Trough. The tectonic plate subduction in the Lesser Antilles, Central America and seafloor spreading is reflected in the morphostructure in the Cayman Trough and Puerto Rico Trench. Modeled cross-sectioning profiles show differences both for the Trench and Trough. In contrast with the Puerto Rico Trench with distinct density peak (680 samples for depths-5,200 to-5,400 m), the Cayman Trough has a bimodal data distribution: two peaks correspond to the two intervals: 1)-3,250 m to-1,000 m; and 2)-5,250 to-3,500. The paper contributes to Caribbean Sea geological studies by using GMT for geomorphological modelling.; W artykule opisano badania wybranych elementów rzeźby dna Morza Karaibskiego na tle budowy geologicznej, tektoniki oraz wybranych właściwości pola geofizycznego wpływające- go na powstawanie rowów głębinowych. Analizę batymetrii wzdłuż Rowu Puerto Rico i Rowu Kajmańskiego wykonano z wykorzystaniem mapowania tematycznego, modelowania geomorfo- logicznego i analizy statystycznej. Wykorzystano zestaw narzędzi do tworzenia skryptów karto- graficznych w środowisku Generic Mapping Tools (GMT). Dane obejmują numeryczny model batymetryczny o strukturze GRID z bazy GEBCO, globalny model geopotencjału EGM2008, anomalie grawitacyjne z uwzględnieniem redukcji wolnopowietrznej (redukcja Faye’a) z połą- czonego zestawu danych altymetrycznych z misji GEOS3/SEASAT/GEOSAT, numeryczny model miąższości osadów o strukturze GRID i rozdzielczości 5 minut z programu GlobSed oraz warstwy wektorowe w formacie GMT (linie brzegowe, sieć rzeczna, granice). Przekrój został wykonany z wykorzystaniem modułu „grdtrack”. Wyniki pokazały różnice między strukturą Rowu Puerto Rico i Rowu Kajmańskiego, na które wpływ miała ewolucja geologiczna. Średnie nachylenie zbo- czy rowu Puerto Rico (67,5°W i 19,90°N do 64,1°W i 19,82°N) wynosi 13°. W części północnej zbocza są bardziej strome (32,09°), ale wyższe na zboczu kontynentalnym. Profile Rowu Puerto Rico są asymetryczne dla obu boków z powodu uskoków i ruchów tektonicznych płyty karaibskiej i płyty północnoamerykańskiej. Dno morskie Rowu Kajmańskiego jest płaskie w tym segmencie (80,0°W i 17,70°N do 78,5°W i 19,50°N). Jego profil jest asymetryczny: północna część jest stro- ma (57°), a południowa jest bardziej łagodna (16°). Bardzo duże ujemne anomalie grawitacyjne Faye’a na wolnym powietrzu (do –380 mGal) widoczne są w Rowie Puerto Rico, na południe od Kuby oraz w północno-wschodniej części Rowu Kajmańskiego. Subdukcja płyt tektonicznych w Małych Antylach, Ameryce Środkowej i na dnie morskim, obejmująca nieckę kajmańską, ko- reluje ze zmianami falowania geoidy wywołanymi właściwościami skał powodujących anomalie grawitacyjne. Analiza warunków topograficznych na przekroju podłużnym ujawnia różnice dla Rowu i niecki. W przeciwieństwie do Rowu Puerto Rico z wyraźnym pikiem gęstości (680 pró- bek dla głębokości od –5200 do –5400 m), Rów Kajmański ma bimodalny rozkład danych: dwa szczyty odpowiadają dwóm interwałom: 1) od –3250 m do –1000 m; 2) od –5250 m do –3500 m. Wnioski zawarte w artykule mogą przyczynić się do badań geologicznych Morza Karaibskiego z technicznym zastosowaniem GMT do modelowania geomorfologicznego.

This research focuses on the 2D and 3D geospatial analysis of the Ryukyu Trench, a deep-sea trench located in the western Pacific Ocean between Japan and Taiwan. The aim of the research is to visualize regional differences in the topography of the southern (S) and northern (N) parts of the trench. Technically, the methodology is based on using the Generic Mapping Tools (GMT) scripting toolset, for modelling the General Bathymetric Chart of the Oceans (GEBCO), and Earth Topography and Bathymetry dataset (ETOPO1) raster grids. The results demonstrated topographic differences in the two segments. The most frequent depths lie between −5,000 and −6,000 m. The N part has steeper gradient slopes and deeper bathymetry. Of the depth differences >−6,000 m, S has nine values with depths >−6,800 m while N shows 123 records (max −7,460 m). The submarine terraces of S have gentler slopes compared with the N segment. The technical approach presents GMT-based 2D and 3D cartographic modelling aimed at visualizing regional variations of the seafloor topography.

This article analyzes topographical and geological settings in the Japan Archipelago for comparative raster data processing using GRASS GIS. Data include bathymetric and geological grids in NetCDF format: GEBCO, EMAG2, GlobSed, marine free‐air gravity anomaly and EGM96. Data were imported to GRASS by gdalwarp utility of GDAL and projected via PROJ library. Method‐ ology includes data processing (projecting and import), mapping and spatial analysis. Visualization was done by shell scripting using a sequence of GRASS modules: ‘d.shade’ for relief mapping, ‘r.slope.aspect’; for modelling based on DEM, ‘r.contour’ for plotting isolines, ‘r.mapcalc’ for classification, ‘r.category’ for associating labels, and auxiliary modules (d.vect, d.rast, d.grid, d.legend). The results of the geophysical visualization show that marine free‐air gravitational anomalies vary in the Sea of Japan (–30 to above 40 mGal) reflecting density inhomogeneities of the tectonic structure, and correlating with the geo‐ logical structure of the seafloor. Dominating values of geoid model are 30–45 m reflecting West Pacific rise, determined by deep density inhomogeneities associated with the mantle convention. Sediment thickness varies across the sea reflecting its geological development with density of 2 km in its deepest part and thinner in central part (Yamato Rise). The aspect map and reclassified map express gradient of the steepest descent.

The paper aims to evaluate the presence and condition of vegetation by SAGA GIS. The study area covers northern coasts of Iceland including two fjords, the Eyjafjörður and the Skagafjörður, prosperous agricultural regions. The vegetation coverage in Iceland experience the impact of harsh climate, land use, livestock grazing, glacial ablation and volcanism. The data include the Landsat TM image. The methodology is based on computing raster bands for simulating Tassel Cap Transformation (wetness, greenness and brightness) and Enhanced Vegetation Index (EVI) sensitive to high biomass. The results include modelled three bands of brightness, greenness and wetness. Greenness variation shows the least values in ice-covered areas (-56.98 to -18.69). High values (-23.48 to 9.12) are in the valleys with dense vegetation, correlating with the geomorphology of the river network, the vegetation-free areas and ocean which corresponds to the peak of 30.87 to 41.19. The bell-shaped data distribution shows frequency 43.19–141.74 for vegetation indicating healthy state and canopy density. Maximal values are in ice-covered regions and glaciers (64°N-65°N). Very low values (0 to -20) show desertification and mountainous rocks. Moderate values (20-40) indicate healthy vegetation. The most frequent data: -28,17 to 11,8. The EVI shows data variations (-0.14 to 0.04). The study contributes both to the regional studies of Arctic Iceland and methodological approach of remote sensing data processing by SAGA GIS.

International audience; The seven-volume set LNCS 12261, 12262, 12263, 12264, 12265, 12266, and 12267 constitutes the refereed proceedings of the 23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020, held in Lima, Peru, in October 2020. The conference was held virtually due to the COVID-19 pandemic.The 542 revised full papers presented were carefully reviewed and selected from 1809 submissions in a double-blind review process. The papers are organized in the following topical sections:Part I: machine learning methodologiesPart II: image reconstruction; prediction and diagnosis; cross-domain methods and reconstruction; domain adaptation; machine learning applications; generative adversarial networksPart III: CAI applications; image registration; instrumentation and surgical phase detection; navigation and visualization; ultrasound imaging; video image analysisPart IV: segmentation; shape models and landmark detectionPart V: biological, optical, microscopic imaging; cell segmentation and stain normalization; histopathology image analysis; opthalmologyPart VI: angiography and vessel analysis; breast imaging; colonoscopy; dermatology; fetal imaging; heart and lung imaging; musculoskeletal imagingPart VII: brain development and atlases; DWI and tractography; functional brain networks; neuroimaging; positron emission tomography

The research problem is about to generate artificial fractal landscape surfaces from the Digital Elevation Model (DEM) using a stochastic algorithm by Geographic Resources Analysis Support System Geographic Information System (GRASS GIS) software. Fractal surfaces resemble appearance of natural topographic terrain and its structure using random surface modelling. Study area covers Kuril-Kamchatka region, Sea of Okhotsk, North Pacific Ocean. Techniques were included into GRASS GIS modules (r.relief, d.rast, r.slope.aspect, r.mapcalc) for raster calculation, processing and visualization. Module 'r.surf.fractal' was applied for generating synthetic fractal surface from ETOPO1 DEM GeoTIFF using algorithm of fractal analysis. Three tested dimensions of the fractal surfaces were automatically mapped and visualized. Algorithm of the automated fractal DEM modelling visualized variations in steepness and aspect of the artificially generated slopes in the mountains. Controllable topographic variation of the fractal surfaces was applied for three dimensions: dim=2.0001, 2.0050, 2.0100. Auxiliary modules were used for the visualization of DEMs (d.rast, r.colors, d.vect, r.contour, d.redraw, d.mon). Modules 'r.surf.gauss' and 'r.surf.random' were applied for artificial modelling as Gauss and random based mathematical surfaces, respectively. Univariate statistics for fractal surfaces were computed for comparative analysis of maps representing continuous fields by module 'r.univar': number of cells, min/max, range, mean, variance, standard deviation, variation coefficient and sum. The paper includes 9 maps and GRASS GIS codes used for visualization., Araştırma problemi, GRASS GIS yazılımı ile stokastik bir algoritma kullanılarak Sayısal Yükseklik Modeli'nden (SYM) yapay fraktal yüzeylerin üretilmesidir. Fraktal yüzeyler, doğal topografik arazinin görünümüne ve yapısına rastgele yüzey modellemesi kullanarak benzerler. Çalışma alanı Kuril-Kamçatka bölgesini, Okhotsk Denizi'ni, Kuzey Pasifik Okyanusu'nu kapsamaktadır. Raster hesaplama, işleme ve görselleştirme için kullanılan yöntemler GRASS GIS modüllerini (r.relief, d.rast, r.slope.aspect, r.mapcalc) içermektedir. Fraktal analiz algoritması kullanılarak ETOPO1 DEM GeoTIFF'den sentetik fraktal yüzey oluşturmak için 'r.surf.fractal' modülü uygulanmıştır. Fraktal yüzeylerin test edilen üç boyutu otomatik olarak haritalanmış ve görselleştirilmiştir. Otomatik fraktal DEM modellemesinin algoritması kullanılarak dağlık alanlarda yapay olarak üretilen yamaçların dikliği ve yönü bakımından oluşturulan varyasyonlar ile görselleştirmeler yapılmıştır. Fraktal yüzeylerin kontrol edilebilir topografik varyasyonu üç boyut için uygulanmıştır: dim = 2.0001, 2.0050, 2.0100. DEM'lerin görüntülenmesi için yardımcı modüller kullanılmıştır (d.rast, r.colors, d.vect, r.contour, d.redraw, d.mon). Yapay modelleme için 'r.surf.gauss' ve 'r.surf.random' modülleri Gauss ve rasgele tabanlı matematiksel yüzeyler olmak üzere sırasıyla uygulanmıştır. Fraktal yüzeyler için tek değişkenli istatistikler 'r.univar' modülüne göre sürekli alanları temsil eden haritaların karşılaştırmalı analizi için hesaplanmıştır: hücre sayısı, min / maks, aralık, ortalama, varyans, standart sapma, varyasyon katsayısı ve toplam. Makalede 9 harita ve görselleştirme için kullanılan GRASS GIS kodları bulunmaktadır.

International audience; This work aim is to perform modelling and spatial analysis of the marine geological data using combination of the QGIS and Python programming. Selecting proper cartographic software is important part of the geospatial research. QGIS provides organizing data in a GIS project for mapping and spatial visualization through vector and raster layers stored in GIS. Study area is Mariana Trench, west Pacific Ocean. A series of cross-section profiles were digitized in QGIS and used for further data processing in Python. Mariana Trench has complex geomorphic structure and unevenness in profiles stretching south-westwards. The geomorphology is subjected to various phenomena that affect its shape. These include bathymetry, geodesy, gravimetry, tectonics plates and geological settings, studied in this paper. To understand the structure of the trench, a data modelling using bathymetric analysis was performed by combination of QGIS mapping and statistical analysis in Python's library Seaborn. Statistical data modelling aimed at the analysis of the spatial variation of the geomorphology of the trench using following methods: multiple facet grids, area charts for the data frames, regression analysis, letter-value plots, hexagonal and Kernel density estimation. The results of the geospatial data analysis show spatial unevenness of the geomorphic structure, gravimetric, geodetic and bathymetric settings of the Mariana Trench. The study demonstrated effectiveness of Python application in geographic data analysis with Python codes provided for repeatability.

We present a method for extracting information about facial expressions from digital images. The method codes facial expression images using a multi-orientation, multi-resolution set of Gabor filters that are topographically ordered and approximately aligned with the face. A similarity space derived from this code is compared with one derived from semantic ratings of the images by human observers. Interestingly the low-dimensional structure of the image-derived similarity space shares organizational features with the circumplex model of affect, suggesting a bridge between categorical and dimensional representations of facial expression. Our results also indicate that it would be possible to construct a facial expression classifier based on a topographically-linked multi-orientation, multi-resolution Gabor coding of the facial images at the input stage. The significant degree of psychological plausibility exhibited by the proposed code may also be useful in the design of human-computer interfaces., Comment: 13 pages, 7 figures, 23 references

The paper presents a comparison of the two languages Python and R related to the classification tools and demonstrates the differences in their syntax and graphical output. It indicates the functionality of R and Python packages {dendextend} and scipy.cluster as effective tools for the dendrogram modelling by the algorithms of sorting and ranking datasets. R and Python programming languages have been tested on a sample dataset including marine geological measurements. The work aims to detect how bathymetric data change along the 25 bathymetric profiles digitized across the Mariana Trench. The methodology includes performed hierarchical cluster analysis with dendrograms and plotted clustermap with marginal dendrograms. The statistical libraries include Matplotlib, SciPy, NumPy, Pandas by Python and {dendextend}, {pvclust}, {magrittr} by R. The dendrograms were compared by the model-simulated clusters of the bathymetric ranges. The results show three distinct groups of the profiles sorted by the elevation ranges with maximal depths detected in a group of profiles 19-21. The dendrogram visualization in a cluster analysis demonstrates the effective representation of the data sorting, grouping and classifying by the machine learning algorithms. The programming codes presented in this study enable to sort a dataset in a similar research aimed to group data based on the similarity of attributes. Effective visualization by dendrograms is a useful modelling tool for the geospatial management where data ranking is required. Plotting dendrograms by R, comparing to Python, presented functional and sophisticated algorithms, refined design control and fine graphical data output. The interdisciplinary nature of this work consists in application of the coding algorithms for spatial data analysis.

CHRIS (Compact High Resolution Imaging Spectrometer) - new imaging spectrometer carried on board a space platform PROBA (Project for On Board Autonomy) The satellite was successfully launched in late October 2001, Shriharikota (India) Region of interest – Thorney Island (UK), Chichester harbour Chichester harbour has unique wetland environment : place for rare bird colonies. Monitoring this place - important ecological task for environmental managers. To obtain a good-quality natural-coloured image of wetlands we need: nadir-taken colour CHRIS image with bands combination of corresponding spectral channels. For the quality of CHRIS images .hdr files were examined: 1) CHRIS...47AO_41.hdr (taken at +36°) 3) CHRIS...47A2_41.hdr (taken at +55°) 2) CHRIS...47A1_41.hdr (taken at -36°) 4) CHRIS...47A3_41.hdr (taken at -55°) 5) CHRIS...479F_41.hdr (taken at nadir) Images taken at the nadir are of good quality, while those at different angles have defects. The work demonstrated CHRIS/PROBA application for Earth Observation purposes, environmental mapping and GIS.

International audience; The current paper deals with the role played by Logarithmic Image Processing (LIP) operators for evaluating the homogeneity of a region. Two new criteria of heterogeneity are introduced, one based on the LIP addition and the other based on the LIP scalar multiplication. Such tools are able to manage Region Growing algorithms following the Revol's technique: starting from an initial seed, they consist of applying specific dilations to the growing region while its inhomogeneity level does not exceed a certain level. The new approaches we introduce are significantly improving Revol's existing technique by making it robust to contrast variations in images. Such a property strongly reduces the chaining effect arising in region growing processes.

International audience; Geometrical and appearance quality requirements set the limits of the current industrial performance in injection molding. To guarantee the product’s quality, it is necessary to adjust the process settings in a closed loop. Those adjustments cannot rely on the final quality because a part takes days to be geometrically stable. Thus, the final part geometry must be predicted from measurements on hot parts. In this paper, we use recent success of Generative Adversarial Networks (GAN) with the pix2pix network architecture to predict the final part geometry, using only hot parts thermographic images, measured right after production. Our dataset is really small, and the GAN learns to translate thermography to geometry. We firstly study prediction performances using different image similarity comparison algorithms. Moreover, we introduce the innovative use of Discrete Modal Decomposition (DMD) to analyze network predictions. The DMD is a geometrical parameterization technique using a modal space projection to geometrically describe surfaces. We study GAN performances to retrieve geometrical parameterization of surfaces.

International audience; Current research presents study of the hydrological characteristics in the surroundings of the Laguna de Gallocanta (Spain) using remote sensing methods and combination of two GIS: ArcGIS and Idrisi GIS. The methodology includes following workflow: 1) selection of the multi-temporal Landsat images and creating color composites; 2) Masking and overlaying study area out of the whole image; 3) Calculation of the direction and length of stream flow in the lake basin; 4) assessment of possible risks of floods and stability of coastal slopes. The research has aim of hydrological assessment testing GIS tools for environmental monitoring. The images are presented by Landsat ETM+, MSS and TM scenes acquired on years 1976, 1987, 1992 and 2000, respectively, with a cloud cover of 0%. The images were georeferenced in ArcGIS and rectified to the Grid Coordinate System using available vector layers. The georeferenced rectified image was then re-projected to the UTM projection WGS84 datum. The combinations of various bands (channels 2-1-4, 5-4-7, and 2-4-7) were tested and visualized. Second step included processing DEM (Digital Elevation Model) Data acquired by the SRTM providing high quality elevation data. The pre-processing included mosaicing images and cutting out study area. The hydrologically conditioning the DEM was performed: the elevations are modified in order to model flow directions. Modelling flow direction and flow accumulation was performed on the conditioned DEM using ArcGIS. The next step included extracting hydrologic features such as stream directions, length and basin extents from a DEM. The pit removal tool was used to remove error pits from a DEM. Final step included calculations of the flow direction and flow length that demonstrate the direction of flow out of each cell. Finally, using Basin function all cells belonging to the watersheds of one or more target cells were calculated. The results shown that area has sensitive hydrological settings which explains its prone to the environmental change. The research demonstrated successful use of GIS for the purposes of hydrological modeling in the environmental studies.

Данная работа представляет результат геопространственного анализа выбранной территории (западная часть Турции, регион Измир) с комбинированным использованием космических снимков Google Earth, Landsat TM и ГИС ПО Erdas Imagine. В работе рассмотрены преимущества совместного использования космических снимков разной категории и детальности (мультиспектральные снимки Landsat TM и ETM+, а также Google Earth веб-сервис) вкупе с ГИС обеспечением для задач геопространственного анализа, часто решаемые в Высшей школе на курсах географии и наук о Земле. Данная работа приводит пример успешного изучения ландшафтов. Изучение распространения различных типов земной поверхности, моделированное с помощью Landsat TM и Google Earth позволяет анализировать динамику изменений ландшафтов, а также подтверждает возможность эффективного использования веб-сервиса Google Earth для тематического картографирования. Для классификации снимков были использованы технологии обработки снимков методами доступных модулей ПО Erdas Imagine. Веб-сервис Google Earth был успешно использован для верификации и валидации результатов картографирования с помощью модуля «Linking with Google Earth».

The research performs urban ecosystem analysis supported by ENVI GIS by integrated studies on land cover types and geospatial modeling of Taipei city. The paper deals with the role of anthropogenic pressure on the structure of the landscape and change of land cover types. Methods included assessment of the impact from anthropogenic activities on the natural ecosystems, evaluation of the rate and scale of landscape dynamics using remote sensing data and GIS. The research aims to assist environmentalists and city planners to evaluate strategies for specific objectives of urban development in Taiwan, China., Comment: 5 pages, 2 figures, 1 table

International audience; The study area is located in Central Europe and focusing on Mecsek Hills, a low mountain range in the south western part of Hungary. The region of central Europe includes complex mis of elements from mediterranean and continental climate, since the area is located in transitional zone of sub-atlantic and sub-mediterranean climate types. Mecsek Hills is a unique region of the Hungarian environment. Its central European location specifies distribution of diverse landscape types, formed under conditions of transitional climatic settings, mixed from sub-atlantic to sub-mediterranean. The distribution of ecosystems in the region illustrate adaptation of soil properties to actual climate conditions. As determined by the present-day climatic and geomorphological settings, soils in the Mecsek region are rich in nutrients, and landscapes are characterized by various mixed soil types. Brown forest soils predominate on hilltops of Mecsek, while alluvial soils on floodplains. According to the underlying soil types, the landscapes are characterized by mixed vegetation types. The most typical is Pannonian mixed forests of temperate broadleaf and mixed forests, common in Euro-Siberian region. There are also forests of Turkey and hornbeam oak in the catchment area of Mecsek region. The dominating vegetation coverage types on the slopes of Mecsek Hills include beech forests, ravine forests and oaks. Some regions of the Mecsek Hills include unique biogeographic areas with endemic species, not founded elsewhere in Carpathian Basin. The overall climate change, especially the global temperature increase, controls balance of heat and water budget on the Hungarian landscapes, and has considerable impact on the agricultural landscapes. In the past decades there were changes in climate, detected in the Carpathian Basin region, which illustrate general fluctuations in the climatic settings in Hungary and change of overall average air temperatures. Combination of the satellite images with GIS techniques is a key method for land patterns identification and classification of ecosystems. Research methodology is based on the GIS spatial analysis tools and classification of Landsat TM image, which was used for visualizing landscapes. An ILWIS GIS was applied to perform spatial analysis and mapping. The research algorithm is clustering, which classifies pixels with similar value of Digital Numbers (DNs) to thematic classes. As a result, spatial analysis on distribution of land cover patterns was performed. Data processing include image pre-processing, enhancement, classification, spatial analysis and interpretation. A land cover map was created by classifying study area into land categories. The core method used in the current work for the interpretation of imagery is clustering algorithm. This method is based on the remote sensing general principle that each unique pixel on a multichannel image has spectral signature defined by the reflectance of its DN in each spectral band. The DNs of pixels create unique signatures for various objects, distinguishable from other objects. Multispectral cluster classification was applied for the Landsat TM image, by extracting information about values of the pixels DNs, analyzing their spectral signatures.Current work demonstrated analysis of climate factors affecting ecological settings of Mecsek Hulls, and local landscapes visualized by means of ILWIS GIS and remote sensing data (Landsat TM). The results consist in recognized distribution of land use patterns. ILWIS GIS is a convenient open source GIS, useful for spatial analysis and land use monitoring. Clustering method is useful for ecological mapping, since it enables objective identification of the land types in regions characterized by high land heterogeneity and complex structure, such as agricultural fields mixed with natural land cover types. The experience of Landsat TM imagery processing by means of ILWIS GIS, described in the current work, is a contribution towards agricultural mapping.

International audience; Asplünd’s metric, which is useful for pattern matching, consists in a double-sided probing, i.e. the over-graph and the sub-graph of a function are probed jointly. It has previously been defined for grey-scale images using the Logarithmic Image Processing (LIP) framework. LIP is a non-linear model to perform operations between images while being consistent with the human visual system. Our contribution consists in extending the Asplünd’s metric to colour and multivariate images using the LIP framework. Asplünd’s metric is insensitive to lighting variations and we propose a colour variant which is robust to noise.

Research area: Taipei urban landscapes, Taiwan. Located on the north of the Taiwan Island, Taipei is the country's core urban political and economic center with population reaching over 2.6 M and continuing to expand within the city Research aim: spatio-temporal analysis of urban dynamics in study area during 15 years (1990-2005) Data capture: The data include Landsat TM images taken from the U.S.G.S. website. The software used is ENVI GIS. Research objective: application of geoinformatic tools, remote sensing data and application of methodology to spatial analysis for a case study of Taipei. Research problem: In the past decades, the process of urbanization has become more notable in the East Asia. Since past decades, Taipei city undergone serious urban changes.The development of urban landscapes in Taipei has unique feature; Taiwan's exceptional economic and industrial growth since 1980s, which is often referred as "industrial wonder of Taiwan". Recently, the country quickly transformed from an agriculture-based traditional economy into a highly industrialized society. As a result, Taiwan has now a new face of high-technology oriented, economically restructured economy. This caused changes in the face of the country, modifying both its social and natural landscape patterns. Taiwan is nowadays a highly urbanized country with almost 80% of the total population of Taiwan living in urban areas. Intensive urbanization necessarily create new conditions for human-nature co-existence, since rapidly growing city and high urbanization rate affect biodiversity and ecosystems sustainability. The process of urbanization includes transformation of natural landscapes: typical natural land cover type are being transformed into human-affected artificial surfaces and the city area enlarges. Anthropogenic pressure increased rapidly due to the expansion of Taipei metropolitan area. Rapid urbanization affects complex interrelations of natural and urban ecosystems, changes their structure, size and shape, which becames a serious environmental problem. At the same time, Taiwan has unique natural landscapes with high environmental value, outstanding natural beauty, rare and extinct species. In vies of this, environmental monitoring should be considered for effective city management and nature conservation on the island. Methods: Workflow includes following steps: Preliminary processing Creation color composites Classification using K-means algorithm Mapping using classification results Accuracy assessment. The preliminary data processing includes image contrast stretching, which is useful as by default, ENVI displays images with a 2% linear contrast stretch. For better contrast the histogram equalization contrast stretch was applied to the image in order to enhance the visual quality. Reults. Based on the image classification, three (3) areas within the city has been selected and distinguished: I, II, III. Comparing regions I, II and III, the conclusion is the following: Region I has changes in the land cover types since 1990s. It has much more changes comparing to the core, old city area (region II) Region II (the core city) is the most stable region. It has the least changes, because this area has been already industrialized for a long time. Region III, has undergone notable changes. This is caused by the intensive relocation of the population to the Taipei area after 1980s. This is regions is dynamically developing within the Taipei since 1990s.

Ce mémoire d'habilitation présente une synthèse de mes réalisations académiques depuis la thèse de doctorat jusqu'à ma récente arrivée à l'IMT Atlantique de Brest. Après une année de post-doctorat en 2011 à l'Université de Technologie du Queensland (QUT, Australie), j'ai effectué la majorité de mes recherches entre 2011 et 2016 au laboratoire d'Ingénierie Circulation Transport (LICIT) de l'IFSTTAR et l'ENTPE de Lyon en tant que chargé de recherches, avant d’intégrer, en 2016, l'IMT Atlantique (ex Télécom Bretagne), au sein de l'équipe DECIDE du lab-STICC (UMR CNRS 6285). Mes thèmes de recherches s’inscrivent dans une application large des mathématiques, à la croisée de l’analyse de données et de la recherche opérationnelle. Pendant les cinq premières années de ma carrière post-thèse, le domaine applicatif de la mobilité intelligente a constitué le fil d’Ariane de mes recherches. Sur la dernière année, mon recrutement à l’IMT Atlantique m’a permis d'approfondir un nouveau champ d’application, la santé connectée, tout en continuant les travaux précédents. Les mutations technologiques ont amené, dans ces deux domaines, une hyper-connectivité, des données ouvertes et de masse (big data), qui ouvrent de formidables perspectives mais aussi une exigence nouvelle de la part des différents acteurs qui souhaitent tout connaître, tout de suite : supervision en temps réel des activités, analyse et prévision des trajets, informations ciblées sur les loisirs, conseils pour leur bien-être, etc. Il s’ensuit un besoin sans précédent de recherche afin de transformer ces masses d’information en connaissances. Au cours des précédentes années, mes travaux ont tenté de contribuer aux domaines disciplinaires de l’informatique et des mathématiques appliquées afin de mieux comprendre et caractériser la mobilité des usagers, ou fournir des outils d’aide à la décision pour la prise en charge de patients à risque en santé mentale. La soutenance d’HDR pose d’abord les motivations de nos travaux sur la mobilité intelligente et donne une présentation originale et synthétique de notre démarche, fondée sur trois directions de recherche. Le cheminement logique entre les trois thèmes de recherche s'appuie sur des montées en échelle successives, du véhicule au réseau, en passant par les données générées par les véhicules. Le premier type de contributions concerne la modélisation et la simulation des véhicules connectés (C-ITS). Ce travail est relatif à la question de l'anticipation des nouvelles technologies, il est centré sur le véhicule à l'aide d'une approche de type model driven. Le deuxième thème répond à la question de l'extraction de connaissances à partir des nouvelles sources de données, centré sur les nouvelles données générées par (entre autres) les nouveaux véhicules à l'aide d'approches de type data driven. Nous présentons quelques résultats sur la caractérisation des états de trafic, à travers la modélisation des temps de parcours et de leur incertitude, l'analyse spatiotemporelle du réseau et la prévision court terme par des méthodes d'apprentissage statistique. Le troisième thème montre comment les résultats des deux premières approches peuvent offrir une meilleure planification des trajets à l'échelle d'un réseau. Il répond à la question de l'optimisation des itinéraires dans les réseaux de transports, il est centré sur l'usager qui souhaite se déplacer d'un point A à un point B et le problème est résolu à l'aide d'algorithmes de recherche de plus court chemin. Dans une deuxième partie, la soutenance d’HDR synthétise nos premiers résultats sur le thème de la santé connectée, résultant de la collaboration avec le CHU Brest. Nous verrons comment des méthodes d'apprentissage non supervisé et supervisé permettent de faire avancer la recherche en santé mentale, par l'extraction de connaissances à partir de grandes bases de données patients. Pour finir, nous inscrirons la mobilité et la santé connectées dans un projet de recherche global pour les années à venir, ayant trait à l'extraction de connaissances et l'optimisation sous contraintes complexes pour des données hétérogènes.

Ce mémoire d'habilitation présente une synthèse de mes réalisations académiques depuis la thèse de doctorat jusqu'à ma récente arrivée à l'IMT Atlantique de Brest. Après une année de post-doctorat en 2011 à l'Université de Technologie du Queensland (QUT, Australie), j'ai effectué la majorité de mes recherches entre 2011 et 2016 au laboratoire d'Ingénierie Circulation Transport (LICIT) de l'IFSTTAR et l'ENTPE de Lyon en tant que chargé de recherches, avant d’intégrer, en 2016, l'IMT Atlantique (ex Télécom Bretagne), au sein de l'équipe DECIDE du lab-STICC (UMR CNRS 6285). Mes thèmes de recherches s’inscrivent dans une application large des mathématiques, à la croisée de l’analyse de données et de la recherche opérationnelle. Pendant les cinq premières années de ma carrière post-thèse, le domaine applicatif de la mobilité intelligente a constitué le fil d’Ariane de mes recherches. Sur la dernière année, mon recrutement à l’IMT Atlantique m’a permis d'approfondir un nouveau champ d’application, la santé connectée, tout en continuant les travaux précédents. Les mutations technologiques ont amené, dans ces deux domaines, une hyper-connectivité, des données ouvertes et de masse (big data), qui ouvrent de formidables perspectives mais aussi une exigence nouvelle de la part des différents acteurs qui souhaitent tout connaître, tout de suite : supervision en temps réel des activités, analyse et prévision des trajets, informations ciblées sur les loisirs, conseils pour leur bien-être, etc. Il s’ensuit un besoin sans précédent de recherche afin de transformer ces masses d’information en connaissances. Au cours des précédentes années, mes travaux ont tenté de contribuer aux domaines disciplinaires de l’informatique et des mathématiques appliquées afin de mieux comprendre et caractériser la mobilité des usagers, ou fournir des outils d’aide à la décision pour la prise en charge de patients à risque en santé mentale. La soutenance d’HDR pose d’abord les motivations de nos travaux sur la mobilité intelligente et donne une présentation originale et synthétique de notre démarche, fondée sur trois directions de recherche. Le cheminement logique entre les trois thèmes de recherche s'appuie sur des montées en échelle successives, du véhicule au réseau, en passant par les données générées par les véhicules. Le premier type de contributions concerne la modélisation et la simulation des véhicules connectés (C-ITS). Ce travail est relatif à la question de l'anticipation des nouvelles technologies, il est centré sur le véhicule à l'aide d'une approche de type model driven. Le deuxième thème répond à la question de l'extraction de connaissances à partir des nouvelles sources de données, centré sur les nouvelles données générées par (entre autres) les nouveaux véhicules à l'aide d'approches de type data driven. Nous présentons quelques résultats sur la caractérisation des états de trafic, à travers la modélisation des temps de parcours et de leur incertitude, l'analyse spatiotemporelle du réseau et la prévision court terme par des méthodes d'apprentissage statistique. Le troisième thème montre comment les résultats des deux premières approches peuvent offrir une meilleure planification des trajets à l'échelle d'un réseau. Il répond à la question de l'optimisation des itinéraires dans les réseaux de transports, il est centré sur l'usager qui souhaite se déplacer d'un point A à un point B et le problème est résolu à l'aide d'algorithmes de recherche de plus court chemin. Dans une deuxième partie, la soutenance d’HDR synthétise nos premiers résultats sur le thème de la santé connectée, résultant de la collaboration avec le CHU Brest. Nous verrons comment des méthodes d'apprentissage non supervisé et supervisé permettent de faire avancer la recherche en santé mentale, par l'extraction de connaissances à partir de grandes bases de données patients. Pour finir, nous inscrirons la mobilité et la santé connectées dans un projet de recherche global pour les années à venir, ayant trait à l'extraction de connaissances et l'optimisation sous contraintes complexes pour des données hétérogènes.

K{\"o}hler's method is a useful multi-thresholding technique based on boundary contrast. However, the direct algorithm has a too high complexity-O(N 2) i.e. quadratic with the pixel numbers N-to process images at a sufficient speed for practical applications. In this paper, a new algorithm to speed up K{\"o}hler's method is introduced with a complexity in O(N M), M is the number of grey levels. The proposed algorithm is designed for parallelisation and vector processing , which are available in current processors, using OpenMP (Open Multi-Processing) and SIMD instructions (Single Instruction on Multiple Data). A fast implementation allows a gain factor of 405 in an image of 18 million pixels and a video processing in real time (gain factor of 96)., Comment: IEEE CopyrightProceedings of the IEEE International Conference on Image Processing ICIP 2017

We introduce a simple expression for the map of Asplund's distances with the multiplicative Logarithmic Image Processing (LIP) law. It is a difference between a morphological dilation and a morphological erosion with an additive structuring function which corresponds to a morphological gradient., Accepted to the 12th European Congress for Stereology and Image Analysis 2017, Kaiserslautern, Germany, September 11-14, 2017

International audience; Predictive models ground many state-of-the-art developments in statistical brain image analysis: decoding, MVPA, searchlight, or extraction of biomarkers. The principled approach to establish their validity and usefulness is cross-validation, testing prediction on unseen data. Here, I would like to raise awareness on error bars of cross-validation, which are often underestimated. Simple experiments show that sample sizes of many neuroimaging studies inherently lead to large error bars, eg ±10% for 100 samples. The standard error across folds strongly underestimates them. These large error bars compromise the reliability of conclusions drawn with predictive models, such as biomarkers or methods developments where, unlike with cognitive neuroimaging MVPA approaches, more samples cannot be acquired by repeating the experiment across many subjects. Solutions to increase sample size must be investigated, tackling possible increases in heterogeneity of the data.

In this paper, we open new perspectives for mathematical expression recognition by introducing an original bimodal system. Since handwritten mathematical expression recognition is a very challenging task prone to many ambiguities, we use speech as an additional modality to circumvent limitations that are inherent to the written form. A use case scenario corresponds to lectures given in classrooms where the teacher would write and read aloud any mathematical expressions to allow a better interpretation. In addition to state-of-the-art solutions for recognizing handwriting and speech, we introduce a multilayer architecture for the merger of modalities. Specifically, the Dempster–Shafer theory is used to process the information at the symbol level. This bimodal system is evaluated on real bimodal data, the HAMEX dataset. Large improvements are observed when speech and handwriting are combined when compared to the single handwriting modality.