Search

International audience; The ANR SO-DRIIHM project (October 2020 - March 2024) has two main objectives, firstly to promote open science within the multidisciplinary DRIIHM community, and secondly to co-design an ergonomic "open science gateway" platform. This presentation provides an overview of the tasks carried out for each work package:- WP1: a survey on data sharing practices, the support of OHM pilot projects (data management plans, data deposits, data paper...) and the search tool for open science resources called ORRSO. We have also contributed to the data policy roadmap for CNRS-INEE infrastructures.- WP2: identify services (search data and publications, deposit data through our platform, make FAIR data as possible, draft a data paper...), set up an IT team with additional funding and carry out IT developments, involve volunteer researchers and the data steward support team in codesigning the future platform.

Iconographic representations, such as historical photos of geographic spaces, are precious cultural heritage resources capable of describing a particular geographical area’s evolution over time. These photographic collections may vary in size, between hundreds and thousands of items. With the advent of the digital era, many of these documents have been digitized, spatialized, and are available online. Browsing through these digital image collections represents new challenges. This paper examines the topic of historical image exploration in a virtual environment enabling the co-visualization of historical photos into a contemporary 3D scene. We address the topic of user interaction considering the potential volume of the input data. Our methodology is based on design guidelines that rely on visual perception techniques to ease visual complexity and improve saliency on specific cues. The designs are additionally implemented following an image-based rendering approach and evaluated in a group of users. Overall, these propositions may be a notable addition to creating innovative ways to visualize and discover historical images in a virtual geographic environment.

In order to understand and explain urban climate, the visual analysis of urban climate data and their relationships with the urban morphology is at stake. This involves partly to co-visualize 3D field climate data, obtained from simulation, with urban 3D models. We propose two ways to visualize and navigate into simulated climate data in urban 3D models, using series of horizontal 2D planes and 3D point clouds. We then explore different parameters regarding transparency, 3D semiologic rules, filtering and animation functions in order to improve the visual analysis of climate data 3D distribution. To achieve this, we apply our propositions to the co-visualization of air temperature data with a 3D urban city model.

Cet article présente quelques illustrations d.applications de données radar polarimétriques (complètes ou partielles) pour le suivi des surfaces terrestres. La première partie est dédiée aux données radar pleinement polarimétriques. Elle est introduite par un rappel théorique présentant leur spécificité. Puis quelques exemples sont présentés, principalement concernant la cartographie de la végétation et également les processus spatio-temporels de sédimentation dans une zone semi-aride en Tunisie. La deuxième partie est consacrée aux données partiellement polarimétriques du type de celles acquises par les capteurs Radar à Synthèse d´Ouverture (RSO) à bord des satellites Sentinel-1A/1B. Celles-ci seront largement utilisés dans les années futures en raison de leur apport significatif pour les études touchant aux observations des surfaces terrestres pour les sciences de l'environnement.

International audience; In many classification tasks, the set of target classes can be organized into a hierarchy. This structure induces a semantic distance between classes, and can be summarized under the form of a cost matrix, which defines a finite metric on the class set. In this paper, we propose to model the hierarchical class structure by integrating this metric in the supervision of a prototypical network. Our method relies on jointly learning a feature-extracting network and a set of class prototypes whose relative arrangement in the embedding space follows an hierarchical metric. We show that this approach allows for a consistent improvement of the error rate weighted by the cost matrix when compared to traditional methods and other prototype-based strategies. Furthermore, when the induced metric contains insight on the data structure, our method improves the overall precision as well. Experiments on four different public datasets-from agricultural time series classification to depth image semantic segmentation-validate our approach.

This article is dedicated to the spatialization of image contents, with a focus on geographical iconographic heritage, i.e. digitized or born-digital image collections, acquired at variable temporal periods and showing the territory and its human-made and natural visual landmarks. We present a panorama of the current solutions (manual, semi-automatic and fully automatic alternatives) that exist to spatialize a visual content, with respect to the data available and the level of spatialization targeted. In particular, we highlight the characteristics of the approaches dedicated to geographical iconographic heritage, and in some cases, we present tests and practical feedbacks that we had the opportunity to conduct for old photographic contents in oblique aerial and terrestrial imagery.

XXIV International Society for Photogrammetry and Remote Sensing Congress, NICE, FRANCE, 04-/07/2021 - 10/07/2021; We present our contribution to the geovisualization and visual analysis of hydraulic simulation data, based on an interdisciplinary research work undertaken by researchers in geographic information sciences and in hydraulics. The positive feedback loop between researchers favored the proposal of visualization tools enabling visual reasoning on hydraulic simulated data so as to infer knowledge on the simulation model. We interactively explore and design 2D multi-scale styles to render hydraulic simulated data, in order to support the identification over large simulation domains of possible local inconsistencies related to input simulation data, simulation parameters or simulation workflow. Models have been implemented into QGIS and are reusable for other input data and territories.

Different forms of phase noise in SAR interferograms hamper the accuracy and reliability of interferometric results (InSAR and DInSAR) for the reconstruction of altimetric information and its variations. Geometric and temporal decorrelations, changes in the properties of observed surfaces and delays introduced by the atmospheric layers are the main sources of interferometric noise. In order to reduce their effects and thus increase the robustness of the phase unwrapping algorithms, filtering methods are applied at different levels of the InSAR and DInSAR processes. In this article, we propose an interferometric filtering process that combines wavelet decomposition into subbands and nonlinear weighting filter called spectral contextual filter (SCF). SCF is based on the Goldstein algorithm, whose nonlinear filter coefficients are calculated, for each element of window according to the filter's punctual parameter. This parameter is determined by estimating the adaptive coherent pseudocorrelation of overlapping blocks of the interferometric phase which combines coherence and pseudocorrelation that are both weighted by the coefficient of variation of each block. In order to show the efficiency and robustness of this process, we compared it with the subspaces decomposition filtering process that we implemented. An assessment scheme was led on the basis of a compromise between residues reduction and fringe boundaries conservation. This assessment was carried relative to the global and local parameters. The tests were conducted on complex interferograms acquired on regions with high topographical variations.

International audience; The increasing availability of large-scale Global Positioning System (GPS) data stemming from in-vehicle embedded terminal devices enables the design of methods deriving road network cartographic information from drivers' recorded traces. Some machine learning approaches have been proposed in the past to train automatic road network map inference, and recently this approach has been successfully extended to infer road attributes as well, such as speed limitation or number of lanes. In this paper, we address the problem of detecting traffic signals from a set of vehicle speed profiles, under a classification perspective. Each data instance is a speed versus distance plot depicting over a hundred profiles on a 100-meter-long road span. We proposed three different ways of deriving features: the first one relies on the raw speed measurements; the second one uses image recognition techniques; and the third one is based on functional data analysis. We input them into most commonly used classification algorithms and a comparative analysis demonstrated that a functional description of speed profiles with wavelet transforms seems to outperform the other approaches with most of the tested classifiers. It also highlighted that Random Forests yield an accurate detection of traffic signals, regardless of the chosen feature extraction method, while keeping a remarkably low confusion rate with stop signs.

Automatic map generalization is a complex task that is still a research problem and requires the development of research prototypes before being usable in productive map processes. In the meantime, reproducible research principles are becoming a standard. Publishing reproducible research means that researchers share their code and their data so that other researchers might be able to reproduce the published experiments, in order to check them, extend them, or compare them to their own experiments. Open source software is a key tool to share code and software, and CartAGen is the first open source research platform that tackles the overall map generalization problem: not only the building blocks that are generalization algorithms, but also methods to chain them, and spatial analysis tools necessary for data enrichment. This paper presents the CartAGen platform, its architecture and its components. The main component of the platform is the implementation of several multi-agent based models of the literature such as AGENT, CartACom, GAEL, CollaGen, or DIOGEN. The paper also explains and discusses different ways, as a researcher, to use or to contribute to CartAGen.

International audience; The automation of map generalization has been keeping researchers in cartography busy for years. Particularly great progress was made in the late 90's with the use of the multi-agent paradigm. Although the current use of automatic processes in some national mapping agencies is a great achievement, there are still many unsolved issues and research seems to stagnate in the recent years. With the success of deep learning in many fields of science, including geographic information science, this paper poses the controversial question of the title: is deep learning the new agent, i.e. the technique that will make generalization research bridge the gap to fully automated generalization processes? The paper neither responds a clear yes nor a clear no but discusses what issues could be tackled with deep learning and what the promising perspectives. Some preliminary experiments with building generalization or data enrichments are presented to support the discussion.

National audience; La visualisation est un mode privilégié de l’interaction des utilisateurs avec l’information géographique, et sa représentation efficace est d’autant plus importante que les données sont massives et hétérogènes et que les utilisateurs et les usages sont variés. Au-delà de la visualisation de données 2D sur un écran ou une carte papier, la visualisation de données 3D pose de nouveaux défis et nécessite des outils appropriés : volume des données, multiplicité des formats, stylisation et gestion des parties visibles, modes d’interaction et de navigation...iTowns est une plateforme technologique de l’IGN qui permet de visualiser des données géographiques 3D via le Web et propose des fonctions d’interaction avancées dans un environnement métrologique.Initialement développé par les laboratoires de recherche de l’IGN comme un outil de visualisation de données images et LiDAR issues de la cartographie mobile (c’est-à-dire acquises au moyen d’un véhicule équipé de capteurs), iTowns a évolué et permet aujourd’hui de naviguer de façon immersive au sein d’un très grand volume de données 3D, et ce dans toute la gamme d’échelles, depuis l’espace jusqu’au sol. Des interfaces sont également disponibles pour la manipulation de ces données. Désormais moteur de visualisation 3D du Géoportail [1], iTowns s’enrichit continuellement de nouvelles fonctionnalités : en sus de la visualisation en 3D du territoire pour le grand public, il permet de développer des applications Web à usage professionnel pour co-visualiser différents types de données, les annoter, procéder à des analyses, des mesures...

For a long time people have been interested in the past and history and how we can go back in time using a time machine. And while we cannot invent a time machine, at least not yet, we can create a virtual one. Our “virtual” time machine is an interactive web application that allows users to browse through and navigate within historical images (aerial/terrestrial photographs or postcards) that are projected on a 3D photogrammetric model (point cloud or 3D mesh), thus going back in time and interacting with historical 3D models and images. This was achieved by adopting a semiautomatic approach where the user identifies first 6 to 8 hints on the historical image and the photogrammetric model, then this information is used as an entry data to a photogrammetric software that computes the pose and orientation of the image. The purpose of this work, which is part of the ALEGORIA project, is to preserve cultural heritage, to give the users the opportunity to go back in time and study history of a place or simply discover how their hometown looked some years ago.

Au milieu des années 1950 surgissent à Nanterre les premières cabanes de fortune abritant des travailleurs algériens. Bientôt, ces baraques informelles s’agrègent et finissent par former des ensembles urbains, présentés et administrés comme des bidonvilles, et la ville de Nanterre y est alors durablement associée. Cet article se propose de revenir sur une expérience de recherche interdisciplinaire autour de cet objet d’étude. Des chercheurs en histoire et en sociologie urbaine ont collaboré avec des informaticiens de l’IGN afin d’utiliser et d’enrichir une plateforme de spatialisation et de visualisation de données hétérogènes pour documenter l’histoire de ces bidonvilles et comprendre la formation et la permanence de ces lieux dans la mémoire collective actuelle. In the mid-1950s, the first improvised huts for Algerian workers appeared in Nanterre. Soon, these informal shacks grew together and ended up forming urban complexes, presented and administered as shantytowns, and the city of Nanterre was then durably associated with them. This article proposes to review an interdisciplinary research experience around this object of study. Researchers in urban history and sociology collaborated with computer scientists from the IGN in order to use and enrich a platform for spatialisation and visualisation of heterogeneous data to document the history of these shantytowns and to understand the formation and permanence of these places in the current collective memory.

Unprecedented access to multi-temporal satellite imagery has opened new perspectives for a variety of Earth observation tasks. Among them, pixel-precise panoptic segmentation of agricultural parcels has major economic and environmental implications. While researchers have explored this problem for single images, we argue that the complex temporal patterns of crop phenology are better addressed with temporal sequences of images. In this paper, we present the first end-to-end, single-stage method for panoptic segmentation of Satellite Image Time Series (SITS). This module can be combined with our novel image sequence encoding network which relies on temporal self-attention to extract rich and adaptive multi-scale spatio-temporal features. We also introduce PASTIS, the first open-access SITS dataset with panoptic annotations. We demonstrate the superiority of our encoder for semantic segmentation against multiple competing architectures, and set up the first state-of-the-art of panoptic segmentation of SITS. Our implementation and PASTIS are publicly available., Comment: Accepted at ICCV2021, PASTIS Dataset available at https://github.com/VSainteuf/pastis-benchmark, PyTorch implementation at https://github.com/VSainteuf/utae-paps

Radar data at C-band has shown great potential for the monitoring of soil and canopy hydric conditions of wheat crops. In this study, the C-band Sentinel-1 time series including the backscattering coefficients u(0) at VV and VH polarization, the polarization ratio (PR) and the interferometric coherence p are first analyzed with the support of experimental data gathered on three plots of irrigated winter wheat located in the Haouz plain in the center of Morocco covering five growing seasons. The results showed that p and PR are tightly related to the canopy development. p is also sensitive to soil preparation. By contrast, u(0) was found to be widely linked to changes in surface soil moisture (SSM) during the first growth stages when Leaf Area Index remains moderate (< 1.5 m(2)/m(2)). In addition, drastic changes in the crop geometry associated to heading had a strong impact on the C-band u(0), in particular for VH polarization. The coupled water cloud and Oh models (WCM) were then calibrated and validated on the study sites. The comparison between the predicted and observed u0 yielded a root mean square error (RMSE) values ranging from 1.50 dB to 2.02 dB for VV and between 1.74 dB to 2.52 dB for VH with significant differences occurring in the second part of the season after heading. Finally, new approaches based on the inversion of the WCM for SSM retrieval over wheat fields were proposed using Sentinel-1 radar data only. To this objective, the dry above-ground biomass (AGB) and the vegetation water content (VWC) were retrieved from the interferometric coherence and the PR. The relationships were then used as the vegetation descriptor in the WCM. The best retrieval results were obtained using the relationship between pVV and the AGB (R and RMSE of 0.82, 0.05 m3/m3 respectively and no bias). The new retrieval approaches were then applied to a large database covering a rainfed field in Morocco and 18 plots of rainfed and irrigated wheat of the Kairouan plain (Tunisia) and compared to other classical techniques of SSM retrieval including simple linear relationships between SSM and u0. The method based on the WCM and the pVV-AGB relationships also provided with slightly better results than the others on the validation database (r = 0.75, RMSE = 0.06 m(3)/m(3) and bias = 0.01 m(3)/m(3) over the 18 plots of Tunisia) but the simple linear relationships performed also reasonably well (r = 0.62, RMSE = 0.07, bias = -0.01 in Tunisia for instance). This study opens perspectives for high resolution soil moisture mapping from Sentinel-1 data over south Mediterranean wheat crops and in fine, for irrigation scheduling and retrieval through the assimilation of these new products in an evapotranspiration model.

Global Navigation Satellite Systems (GNSS) have been used in various monitoring applications for the past two decades, as one of the very few options to provide absolute positions in a global refer...

This article presents a spatio-temporal web application dedicated to the co-exploitation of heterogeneous data spatialized in a common 3D environment, providing several paradigms for supporting their co-visualization and interactions within the 3D environment and across time. The relevance of this tool is demonstrated here with two use cases involving historians and sociologists with the common objective of better understanding the formation of the Parisian metropolis. The study focuses on the evolution of the city of Nanterre (Paris area), which underwent many changes in the 1950s, and in particular on shantytown areas. Through census as statistical data and aerial imagery as visual data, a group of historians and sociologists experimented the relevance of the joint exploitation of those heterogeneous data within the proposed spatio-temporal web application.

Archivists, historians and national mapping agencies, among others, are archiving large datasets of historical photographs. Nevertheless, the capturing devices used to acquire these images possessed a diversity of effects that influenced the quality of the final resulting picture, e.g. geometric distortion, chromatic aberration, depth of field variation, etc. This paper examines singularly the topic of geometric distortion for a co-visualization of historical photos within a 3D model of the photographed scene. A distortion function of an image is ordinarily estimated only on the image domain by adjusting its parameters to observations of point correspondences. This mathematical function may exhibit overfits, oscillations or may not be well defined outside of this domain. The contribution of this work is the description of a distortion model defined on the whole undistorted image plane. We extrapolate the distortion estimated only on the image domain and then transfer this distortion information to the view of the 3D scene. This enables to look at the scene through an estimated camera and zoom out to see the context around the original photograph with a well-defined and behaved distortion. These findings may be a significant addition to the overall purpose of creating innovative ways to examine and visualize old photographs.

Though Volunteered Geographic Information (VGI) has the advantage of providing free open spatial data, it is prone to vandalism, which may heavily decrease the quality of these data. Therefore, detecting vandalism in VGI may constitute a first way of assessing the data in order to improve their quality. This article explores the ability of supervised machine learning approaches to detect vandalism in OpenStreetMap (OSM) in an automated way. For this purpose, our work includes the construction of a corpus of vandalism data, given that no OSM vandalism corpus is available so far. Then, we investigate the ability of random forest methods to detect vandalism on the created corpus. Experimental results show that random forest classifiers perform well in detecting vandalism in the same geographical regions that were used for training the model and has more issues with vandalism detection in &ldquo, unfamiliar regions&rdquo

Augmented reality (AR) enables to display situated geographical visualizations, i.e visualizations that use virtual elements that are displayed in a geographical location. The place where the data is displayed complements the visualization. Many applications that take advantage of AR and situated visualizations exist, but they differ in the visualizations they present, their relationship to the geographic locations and goals. To better understand why and how AR based situated geovisualization is used, we review 45 papers coming from Human Computer Interaction, Visualization and Geographical Information Science venues that present such applications. Inspired by existing classifications, we characterize these papers according to the data they visualize and the geographical distance between the visualization and the data the visualization represents. This analysis reveals existing opportunities for situated geovisualization applications using AR.

This manuscript identifies and documents unsolved problems and research challenges in the extended reality (XR) domain (i.e., virtual (VR), augmented (AR), and mixed reality (MR)). The manuscript is structured to include technology, design, and human factor perspectives. The text is visualization/display-focused, that is, other modalities such as audio, haptic, smell, and touch, while important for XR, are beyond the scope of this paper. We further narrow our focus to mainly geospatial research, with necessary deviations to other domains where these technologies are widely researched. The main objective of the study is to provide an overview of broader research challenges and directions in XR, especially in spatial sciences. Aside from the research challenges identified based on a comprehensive literature review, we provide case studies with original results from our own studies in each section as examples to demonstrate the relevance of the challenges in the current research. We believe that this paper will be of relevance to anyone who has scientific interest in extended reality, and/or uses these systems in their research.

This paper introduces a multi-modal network that learns to match aerial images of french urban and rural territories taken about 15 years apart. This means it should be invariant against a big range of changes as the (natural) landscape evolves over time. It leverages the original images and semantically segmented and labeled regions. The core of the method is a siamese network that learns to extract features from corresponding image pairs across time, and non matching pairs. These descriptors are discri-minative enough, such that a simple kNN classifier on top, suffices as a final geo-matching criteria. We demonstrate that our siamese descriptor outperforms other image descriptors for cross-time image retrieval., Cet article présente un réseau multimodal qui met en cor-respondance des images aériennes de territoires urbains et ruraux français prises à environ 15 ans d'intervalle. Il de-vrait être invariant à un large éventail de changements, tels que l'évolution du paysage au fil des années. Il exploite les images originales et les régions sémantiquement segmen-tées et étiquetées. Le coeur de la méthode est un réseau siamois qui apprend à extraire des caractéristiques des paires d'images cor-respondantes dans le temps et des paires non correspon-dantes. Ces descripteurs sont suffisamment discriminants pour qu'un simple classifieur k-NN suffise comme critère de géo-correspondance final. Dans cet article, nous dé-montrons que notre descripteur siamois surpasse les autres descripteurs d'images en termes de recherche d'images par contenu à travers le temps.

Thanks to their ever improving resolution and accessibility, Light Detection And Ranging (LiDAR) sensors are increasingly used for mapping cities. Indeed, these sensors are able to efficiently capture high-density scans, which can then be used to produce geometrically detailed reconstructions of complex scenes. However, such reconstruction requires organizing the scan with a fitting data structure, such as point clouds or meshes. Point clouds provide such a representation in a compact way, but their discrete nature prevents some applications such as visualization or simulation. Meshes allow for a continuous representation of surfaces, but are not well suited for representing complex objects, whose level of detail can exceed the resolution. To address these limitations, we propose to reconstruct a continuous geometry only where sufficient geometric information is available. This leads us to create a reconstruction mixing triangles, edges and points. We call such collection of objects a simplicial complex. In this thesis, we study the creation of geometrically detailed 3-dimensional (3D) models of urban scenes, based on simplicial complexes. We show that simplicial complexes are a suitable alternative to meshes. Indeed, they are fast to compute, and can be simplified while maintaining high geometric geometric fidelity with respect to the input scan. We argue that simplicial complexes convey valuable geometric information which can in turn be used for the semantization of 3D point clouds. We also think that they can serve as input for multi-scale reconstructions of urban scenes. We first present an efficient algorithm for computing simplicial complexes from LiDAR scans of urban scenes. Since the reconstructed simplicial complexes can be very large, they can be difficult to process on a standard computer. To handle this challenge, we investigate different approaches for their spatial generalization by approximating large and geometrically simple areas with simple primitives. To this end, we propose a new algorithm to compute piecewise-planar approximations of 3D point clouds, based on a global optimization approach. Next, we propose two different applications of simplicial complexes. The first one is a polygonalization method improving the creation of light yet geometrically accurate 3D models. The second one is a weakly-supervisedclassification method using 3D local and global descriptors, Grâce à leur résolution et à leur accessibilité toujours meilleures, les capteurs LiDAR sont de plus en plus utilisés pour cartographier les villes. En effet, ces capteurs sont capables de réaliser efficacement des acquisitions à haut résolution, qui peuvent ensuite être utilisées pour produire des reconstructions géométriquement détaillées de scènes complexes. Cependant, une telle reconstruction nécessite d’organiser les données avec une structure de données adaptée, comme des nuages de points ou des maillages. Les nuages de points fournissent une représentation compacte des données, mais leur nature discrète empêche certaines applications telles que la visualisation ou la simulation. Les maillages permettent une représentation continue des surfaces, mais ne sont pas bien adaptés à la représentation d’objets complexes, dont le niveau de détail peut dépasser la résolution de l’acquisition. Pour remédier à ces limitations, nous proposons de reconstruire une géométrie continue uniquement lorsque suffisamment d’informations géométriques sont disponibles. Cela nous amène à créer une reconstruction mêlant triangles, arêtes et points. Nous appelons une telle collection d’objets un complexe simplicial. Dans cette thèse, nous étudions la création de modèles 3D de scènes urbaines géométriquement détaillés, basés sur des complexes simpliciaux. Nous montrons que les complexes simpliciaux sont une alternative appropriée aux maillages. En effet, ils sont rapides à calculer et peuvent être simplifiés tout en conservant une grande fidélité géométrique par rapport aux données d’entrée. Nous soutenons que les complexes simples transmettent de précieuses informations géométriques qui peuvent à leur tour être utilisées pour la sémantisation des nuages de points 3D. Nous pensons également qu’ils peuvent servir de base pour des reconstructions multi-échelles de scènes urbaines. Nous présentons d’abord un algorithme efficace pour le calcul de complexes simpliciaux à partir d’acquisitions LiDAR de scènes urbaines. Comme les complexes simpliciaux reconstruits peuvent être très lourds, ils peuvent être difficiles à traiter sur un ordinateur standard. Pour relever ce défi, nous étudions différentes approches pour les généraliser spatialement, en approximant de grandes zones géométriquement simples par des primitives simples. À cette fin, nous proposons un nouvel algorithme pour calculer des approximations planaires par morceaux de nuages de points 3D, basé sur une approche d’optimisation globale. Ensuite, nous proposons deux applications différentes des complexes simpliciaux. La première est une méthode de polygonalisation améliorant la création de modèles 3D légers mais géométriquement précis. La seconde est une méthode de classification faiblement supervisée utilisant des descripteurs 3D locaux et globaux

International audience; Satellite image time series, bolstered by their growing availability , are at the forefront of an extensive effort towards automated Earth monitoring by international institutions. In particular, large-scale control of agricultural parcels is an issue of major political and economic importance. In this regard, hybrid convolutional-recurrent neural architectures have shown promising results for the automated classification of satellite image time series.We propose an alternative approach in which the convolutional layers are advantageously replaced with encoders operating on unordered sets of pixels to exploit the typically coarse resolution of publicly available satellite images. We also propose to extract temporal features using a bespoke neural architecture based on self-attention instead of recurrent networks. We demonstrate experimentally that our method not only out-performs previous state-of-the-art approaches in terms of precision, but also significantly decreases processing time and memory requirements. Lastly, we release a large open-access annotated dataset as a benchmark for future work on satellite image time series.

Visual localization is a useful alternative to standard localization techniques. It works by utilizing cameras. In a typical scenario, features are extracted from captured images and compared with geo-referenced databases. Location information is then inferred from the matching results. Conventional schemes mainly use low-level visual features. These approaches offer good accuracy but suffer from scalability issues. In order to assist localization in large urban areas, this work explores a different path by utilizing high-level semantic information. It is found that object information in a street view can facilitate localization. A novel descriptor scheme called "semantic signature" is proposed to summarize this information. A semantic signature consists of type and angle information of visible objects at a spatial location. Several metrics and protocols are proposed for signature comparison and retrieval. They illustrate different trade-offs between accuracy and complexity. Extensive simulation results confirm the potential of the proposed scheme in large-scale applications. This paper is an extended version of a conference paper in CBMI'18. A more efficient retrieval protocol is presented with additional experiment results., Comment: 12 pages, 22 figures, Multimedia Tools and Applications (2020)

To preserve and restore ecosystems, public environmental policies on the international level now encourage the creation of Green Infrastructure, i.e. networks composed of areas where animal and veg...

Land use and land cover (LULC) mapping is often undertaken by national mapping agencies, where these LULC products are used for different types of monitoring and reporting applications. Updating of LULC databases is often done on a multi-year cycle due to the high costs involved, so changes are only detected when mapping exercises are repeated. Consequently, the information on LULC can quickly become outdated and hence may be incorrect in some areas. In the current era of big data and Earth observation, change detection algorithms can be used to identify changes in urban areas, which can then be used to automatically update LULC databases on a more continuous basis. However, the change detection algorithm must be validated before the changes can be committed to authoritative databases such as those produced by national mapping agencies. This paper outlines a change detection algorithm for identifying construction sites, which represent ongoing changes in LU, developed in the framework of the LandSense project. We then use volunteered geographic information (VGI) captured through the use of mapathons from a range of different groups of contributors to validate these changes. In total, 105 contributors were involved in the mapathons, producing a total of 2778 observations. The 105 contributors were grouped according to six different user-profiles and were analyzed to understand the impact of the experience of the users on the accuracy assessment. Overall, the results show that the change detection algorithm is able to identify changes in residential land use to an adequate level of accuracy (85%) but changes in infrastructure and industrial sites had lower accuracies (57% and 75 %, respectively), requiring further improvements. In terms of user profiles, the experts in LULC from local authorities, researchers in LULC at the French national mapping agency (IGN), and first-year students with a basic knowledge of geographic information systems had the highest overall accuracies (86.2%, 93.2%, and 85.2%, respectively). Differences in how the users approach the task also emerged, e.g., local authorities used knowledge and context to try to identify types of change while those with no knowledge of LULC (i.e., normal citizens) were quicker to choose ‘Unknown’ when the visual interpretation of a class was more difficult.

International audience; Cities are particularly vulnerable to meteorological hazards because of the concentration of population, goods, capital stock and infrastructure. Urban climate services require multi-disciplinary and multi-sectorial approaches and new paradigms in urban climate modelling. This paper classifies the required urban input data for both mesoscale state-of-the-art Urban Canopy Models (UCMs) and microscale Obstacle Resolving Models (ORM) into five categories and reviews the ways in which they can be obtained. The first two categories are (1) land cover, and (2) building morphology. These govern the main interactions between the city and the urban climate and the Urban Heat Island. Interdependence between morphological parameters and UCM geometric hypotheses are discussed. Building height, plan and wall area densities are recommended as the main input variables for UCMs, whereas ORMs require 3D building data. Recently, three other categories of urban data became relevant for finer urban studies and adaptation to climate change: (3) building design and architecture, (4) building use, anthropogenic heat and socio-economic data, and (5) urban vegetation data. Several methods for acquiring spatial information are reviewed, including remote sensing, geographic information system (GIS) processing from administrative cadasters, expert knowledge and crowdsourcing. Data availability, data harmonization, costs/efficiency trade-offs and future challenges are then discussed.

The use of consumer grade unmanned aerial vehicles (UAV) is becoming more and more ubiquitous in photogrammetric applications. A large proportion of consumer grade UAVs are equipped with CMOS image sensor and rolling shutter. When imaging with a rolling shutter camera, the image sensor is exposed line by line, which can introduce additional distortions in image space since the UAV navigates at a relatively high speed during aerial acquisitions. In this paper, we propose (1) an approach to calibrate the readout time of rolling shutter camera, (2) a two-step method to correct the image distortion introduced by this effect. The two-step method makes assumption that during exposure, the change of camera orientation is negligible with respect to the change of camera position, which is often the case when camera is fixed on a stabilized mount. Firstly, the camera velocity is estimated from the results of an initial bundle block adjustment; then, one camera pose per scan-line of the image sensor is recovered and image observations are corrected. To evaluate the performance of the proposed method, four datasets of block and corridor configurations are acquired with the DJI Mavic 2 Pro and its original Hasselbald L1D-20c camera. The proposed method is implemented in MicMac, a free, open-source photogrammetric software; comparisons are carried out with other two mainstream software, AgiSoft MetaShape and Pix4D, which also have the functionality of rolling shutter effect correction. For block configuration datasets, the three software give comparable results. AgiSoft Metashape and Pix4D are sensitive to the flight configuration and encounter difficulties when processing datasets in corridor configurations. The proposed method shows good robustness both in block and corridor configurations, and is the only method that works in corridor configuration. After the application of the rolling shutter effect correction, the 3D accuracy is improved by 30–60% in block configuration and 15–25% in corridor configuration. A further improvement can be expected if a precise dating of image is available or if the camera positions can be directly extracted from GNSS data.

For political or environmental matters, land cover mapping has become more and more important since the beginning of the XXIst century. This mapping consists in, given a set of land cover classes, classifying ground-level objects according to these classes. Adressing very various applications, from a local scale (city-wise) to a more global scale (world-wide), several projects have been initiated so as to build land cover maps that would fit one of these applications. However, the current means to achieve this mapping are mainly manual, leading to high cost of production both in terms of man power and processing time. The task of building relevant land cover maps lies also in being able to monitor various phenomena in time, but acquiring data can be expansive for frequent updates or computation of maps at large scale. Yet, remote sensors have multiplied for the past two decades and the panel of such sensors is very wide and diverse in a spectral way and in terms of resolution. Among them, optical sensors boarded on Earth observation satellites acquire images in the visible and infrared spectrums, with high revisit time capabilities. SPOT 6 and 7 were launched in 2012 and 2014 respectively, and offer very high resolution images at 1.5m, in four bands. The French Mapping Agency computes each year a full France coverage from the SPOT data available through the THEIA data portal. Although this task has been widely studied in the past by the remote sensing community, by the means of supervised classifiers such as SVMs or random forests, none of the results have yet led to truly automated and satisfying maps that match existing specifications due to mislabelling errors that occur too often. In terms of methodology, the deep learning, and more specifically deep neural networks have proved to be far more efficient than other machine learning algorithms in various fields of research, from computer vision to natural language processing and system recommendation. These days, deep learning has become a part of everyday life, not only for academic research purposes, but also it society where it runs in the background of our smartphones for various tasks. Such ubiquity raises ethical issues: in particular, users’ browsing habits and behaviours are often monitored without them knowing to feed these algorithms. This heavy need for data is what allows deep learning to sketch more accurately a user behaviour, thus leading to more suitable commercial offers based on a history of purchases for instance.In our work, the French Mapping Agency topographic databases represent a real opportunity to massively train deep neural networks on the SPOT monoscopic very high resolution images, so as to produce land cover maps. The works reported in this dissertation are focused on this problematic of land cover mapping with deep neural networks on SPOT 6 and 7 data, with the additional constraint of tackling this task in a more general operational environment (large scale land cover mapping). This latter point ensures that the conducted experiments address occurring questions when classifying large geographic areas. For instance, a specific SPOT coverage is made of one image per territory, which means adjacent areas can have been acquired at different dates, leading to strong appearance discrepancies. Transfer learning is a tool that can help a lot in lots of scenarii so as to reduce both time computation and needs for training samples. Finally, in a context of the automatic update or re-computation of existing topographic database, the use of aerial images in deep neural networks is assessed in several tasks, with an emphasis on the training dataset conception from topographic database that can include some drawbacks, La connaissance de la couverture des territoires en terme d’occupation des sols est devenue un enjeu majeur du XXIème siècle. Que ce soit à l’échelle nationale ou à une échelle plus globale, les initiatives se multiplient pour proposer des cartographies d’occupation des sols qui répondent à des besoins propres à chacune. Consistant à classer des objets présents sur le sol selon des nomenclatures prédéfinies, la tâche est fastidieuse à l’heure actuelle avec des processus essentiellement manuels ou semi-manuels, nécessaires pour garantir le respect de certaines qualités et spécifications. La télédétection spatiale a connu un essor conséquent avec la multiplication des capteurs optiques d’observation de la Terre disponibles et de leur diversité en terme de résolutions spectrale, spatiale et temporelle. Ces capteurs optiques proposent chacun une description de la surface terrestre qui leur est propre. Les satellites SPOT 6 et SPOT 7 sont dotés de capteurs optiques à très haute résolution spatiale, et acquièrent des images dans quatre bandes spectrales à très haute résolution après pan-sharpening (1,5m). L’IGN, à partir de ces acquisitions SPOT disponibles sur le pôle de données surfaces continentales THEIA, produit chaque année une couverture d’orthophotos sur l’ensemble du territoire français. La problématique de cartographie de l’occupation des sols automatique à partir d’images aériennes ou satellites occupe la communauté de télédétection depuis longtemps, par le biais de processus de classification supervisés, tels que les SVMs, ou les forêts aléatoires pour, entre autres, la vitesse d’exécution de ces derniers. Mais les résultats obtenus par ces méthodes n'ont pas encore permis une réelle automatisation, notamment en adéquation avec des spécifications existantes (erreurs encore trop importantes). En parallèle de ces algorithmes depuis longtemps utilisés, des méthodes d’apprentissage automatique d’un genre nouveau, bien que reposant sur des concepts remontant aux années 1950, émergent depuis une décennie. L’apprentissage profond a fait ses preuves dans de nombreux domaines depuis le traitement naturel du langage, à la reconnaissance d’objets dans des images. Cet essor récent est la conséquence de la disponibilité de grandes bases de données d'apprentissage, ainsi que la croissance globale des puissances de calcul. Représentants principaux de cette famille d’apprentissage, les réseaux de neurones profonds ont un impact au quotidien. Que ce soit au niveau académique en terme de recherche, au niveau sociétal, au travers des smartphones par exemple, ou même au niveau politique, avec les questions déontologiques que cela peut poser en terme de confidentialité des données (RGPD) et de protection des libertés individuelles, l'apprentissage profond est au cœur de technologies utilisées par la plupart des gens à leur insu. En effet, pour afficher de telles performances dans tant de domaines, ces algorithmes nécessitent de gros volumes de données. Les bases de données géographiques de l’IGN sont donc une opportunité dans notre cas, permettant d’exploiter au mieux les images très haute résolution monoscopiques acquises par les satellites SPOT 6 et 7 en les classifiant automatiquement par réseaux de neurones profonds appris sur ces mêmes bases de données. C’est cette approche que nous proposons dans ces travaux de thèse, avec une volonté d’étudier cette problématique tout en se plaçant dans un cadre plus large à visée opérationnelle. Nous étudions les possibilités de transfert d’apprentissage qui offre beaucoup d’avantages en matière de charges de calcul et de jeu d’apprentissage. Enfin, l’exploitation jointe d’images aériennes et de réseaux de neurones profonds est étudiée, avec un accent mis sur la préparation des données d’apprentissage issues des bases de données géographiques de l’IGN qui présentent certains inconvénients

The quality of Volunteered Geographic Information (VGI) is currently a topic that question spatial data users as well as authoritative data producers who are willing to exploit the benefits of crowdsourcing. Contrary to most authoritative databases, the advantage of VGI provides open access to spatial data. However, VGI is prone to errors, even to deliberate defacement perpetrated by ill-intended contributors. In the latter case, we may speak of cartographic vandalism of carto-vandalism. This phenomenon is one the main downsides of crowsdsourcing, and despite the small amount of incidents, it may be a barrier to the use of collaborative spatial data. This thesis follows an approach based on VGI quality -- in particular, the objective of this work is to detect vandalism in spatial collaborative data. First, we formalize a definition of the concept of carto-vandalism. Then, assuming that corrupted spatial data come from malicious contributors, we demonstate that qualifying contributors enables to assess the corresponding contributed data. Finally, the experiments explore the ability of learning methods to detect carto-vandalism; La qualité de l'information géographique volontaire est actuellement un sujet qui questionne autant les consommateurs de données géographiques que les producteurs de données d'autorité voulant exploiter les bienfaits de la démarche collaborative. En effet, si les données cartographiques collaboratives présentent l'intérêt d'être ouvertes, contrairement à certaines bases de données géographiques officielles, celles-ci sont néanmoins sujettes à des erreurs voire à des dégradations volontaires, provoquées par des contributeurs malintentionnés. Dans ce dernier cas, on parle de vandalisme cartographique ou de carto-vandalisme. Ce phénomène est un travers de la démarche collaborative, et bien qu'il ne concerne qu'une faible portion des contributions, il peut constituer un obstacle à l'utilisation des données cartographiques participatives. Dans une démarche de qualification de l'information géographique volontaire, ce travail de thèse a plus précisément pour objectif de détecter le vandalisme dans les données collaboratives cartographiques. Dans un premier temps, il s'agit de formaliser une définition du concept de carto-vandalisme. Puis, en partant du principe que les contributions volontairement dégradées proviennent de contributeurs malveillants, nous cherchons à démontrer que la qualification des contributeurs permettant d'évaluer leurs contributions. Enfin, nos expériences explorent la capacité des méthodes d'apprentissage machine (machine learning) à détecter le carto-vandalisme

International audience; The relation between slip at the near surface and at depth during earthquakes is still not fully resolved at the moment. This deficiency leads to large uncertainties in the evaluation of the magnitude of past earthquakes based on surface observations, which is the only accessible evidence for such events. A better knowledge of the way slip distributes over distinct rupture strands within the first few kilometers from the surface would contribute greatly to reduce these uncertainties. The 30 October 2016 Mw6.5 Norcia earthquake has been captured by a variety of geodetic techniques, which provide access to the slip distribution both at depth and at the ground surface, with an unprecedented level of detail for a normal‐faulting earthquake. We first present coseismic surface offset measurements from correlation of optical satellite images of submetric resolution, which are compared to field observations made shortly after the earthquake. Based on a joint inversion of optical data together with InSAR and GPS data, we then propose a rupture model that explains the observations both at far‐field and near‐field scales. Finally, we explore different rupture geometries at shallow depth, in an attempt to better explain the near‐field deformation (i.e., within the first hundreds of meters around the fault) observed at the surface. Despite the fact that the solution is not unique, several lines of evidence suggest that gravity processes could be locally involved, which interfere with the dominant tectonic processes.

International audience

International audience

International audience; Leveraging the recent availability of accurate, frequent, and multimodal (radar and optical) Sentinel-1 and-2 acquisitions, this paper investigates the automation of land parcel identification system (LPIS) crop type classification. Our approach allows for the automatic integration of temporal knowledge, i.e., crop rotations using existing parcel-based land cover databases and multi-modal Sentinel-1 and-2 time series. The temporal evolution of crop types was modeled with a linearchain conditional random field, trained with time series of multi-modal (radar and optical) satellite acquisitions and associated LPIS. Our model was tested on two study areas in France (≥ 1250 km 2) which show different crop types, various parcel sizes, and agricultural practices: .the Seine et Marne and the Alpes de Haute-Provence classified accordingly to a fine national 25-class nomenclature. We first trained a Random Forest classifier without temporal structure to achieve 89.0% overall accuracy in Seine et Marne (10 classes) and 73% in Alpes de Haute-Provence (14 classes). We then demonstrated experimentally that taking into account the temporal structure of crop rotation with our model resulted in an increase of 3% to +5% in accuracy. This increase was especially important (+12%) for classes which were poorly classified without using the temporal structure. A stark positive impact was also demonstrated on permanent crops, while it was fairly limited or even detrimental for annual crops. Crop Rotation Integration Crop rotation knowledge can be used to improve agricultural yields (Berzsenyi, Györffy, and Lap et al. 2000) and soil quality (Karlen, Hurley, Andrews et al. 2006). Crop type prediction can also be improved using prior knowledge on crop rotations per parcel since a crop type is strongly correlated to past crop types. Modeling such temporal structures from

Spatial analysis and pattern recognition with vector spatial data is particularly useful to enrich raw data. In road networks, for instance, there are many patterns and structures that are implicit with only road line features, among which highway interchange appeared very complex to recognize with vector-based techniques. The goal is to find the roads that belong to an interchange, such as the slip roads and the highway roads connected to the slip roads. To go further than state-of-the-art vector-based techniques, this article proposes to use raster-based deep learning techniques to recognize highway interchanges. The contribution of this work is to study how to optimally convert vector data into small images suitable for state-of-the-art deep learning models. Image classification with a convolutional neural network (i.e., is there an interchange in this image or not?) and image segmentation with a u-net (i.e., find the pixels that cover the interchange) are experimented and give better results than existing vector-based techniques in this specific use case (99.5% against 74%).

International audience

National audience

For political or environmental matters, land cover mapping has become more and more important since the beginning of the XXIst century. This mapping consists in, given a set of land cover classes, classifying ground-level objects according to these classes. Adressing very various applications, from a local scale (city-wise) to a more global scale (world-wide), several projects have been initiated so as to build land cover maps that would fit one of these applications. However, the current means to achieve this mapping are mainly manual, leading to high cost of production both in terms of man power and processing time. The task of building relevant land cover maps lies also in being able to monitor various phenomena in time, but acquiring data can be expansive for frequent updates or computation of maps at large scale. Yet, remote sensors have multiplied for the past two decades and the panel of such sensors is very wide and diverse in a spectral way and in terms of resolution. Among them, optical sensors boarded on Earth observation satellites acquire images in the visible and infrared spectrums, with high revisit time capabilities. SPOT 6 and 7 were launched in 2012 and 2014 respectively, and offer very high resolution images at 1.5m, in four bands. The French Mapping Agency computes each year a full France coverage from the SPOT data available through the THEIA data portal. Although this task has been widely studied in the past by the remote sensing community, by the means of supervised classifiers such as SVMs or random forests, none of the results have yet led to truly automated and satisfying maps that match existing specifications due to mislabelling errors that occur too often. In terms of methodology, the deep learning, and more specifically deep neural networks have proved to be far more efficient than other machine learning algorithms in various fields of research, from computer vision to natural language processing and system recommendation. These days, deep learning has become a part of everyday life, not only for academic research purposes, but also it society where it runs in the background of our smartphones for various tasks. Such ubiquity raises ethical issues: in particular, users’ browsing habits and behaviours are often monitored without them knowing to feed these algorithms. This heavy need for data is what allows deep learning to sketch more accurately a user behaviour, thus leading to more suitable commercial offers based on a history of purchases for instance.In our work, the French Mapping Agency topographic databases represent a real opportunity to massively train deep neural networks on the SPOT monoscopic very high resolution images, so as to produce land cover maps. The works reported in this dissertation are focused on this problematic of land cover mapping with deep neural networks on SPOT 6 and 7 data, with the additional constraint of tackling this task in a more general operational environment (large scale land cover mapping). This latter point ensures that the conducted experiments address occurring questions when classifying large geographic areas. For instance, a specific SPOT coverage is made of one image per territory, which means adjacent areas can have been acquired at different dates, leading to strong appearance discrepancies. Transfer learning is a tool that can help a lot in lots of scenarii so as to reduce both time computation and needs for training samples. Finally, in a context of the automatic update or re-computation of existing topographic database, the use of aerial images in deep neural networks is assessed in several tasks, with an emphasis on the training dataset conception from topographic database that can include some drawbacks, La connaissance de la couverture des territoires en terme d’occupation des sols est devenue un enjeu majeur du XXIème siècle. Que ce soit à l’échelle nationale ou à une échelle plus globale, les initiatives se multiplient pour proposer des cartographies d’occupation des sols qui répondent à des besoins propres à chacune. Consistant à classer des objets présents sur le sol selon des nomenclatures prédéfinies, la tâche est fastidieuse à l’heure actuelle avec des processus essentiellement manuels ou semi-manuels, nécessaires pour garantir le respect de certaines qualités et spécifications. La télédétection spatiale a connu un essor conséquent avec la multiplication des capteurs optiques d’observation de la Terre disponibles et de leur diversité en terme de résolutions spectrale, spatiale et temporelle. Ces capteurs optiques proposent chacun une description de la surface terrestre qui leur est propre. Les satellites SPOT 6 et SPOT 7 sont dotés de capteurs optiques à très haute résolution spatiale, et acquièrent des images dans quatre bandes spectrales à très haute résolution après pan-sharpening (1,5m). L’IGN, à partir de ces acquisitions SPOT disponibles sur le pôle de données surfaces continentales THEIA, produit chaque année une couverture d’orthophotos sur l’ensemble du territoire français. La problématique de cartographie de l’occupation des sols automatique à partir d’images aériennes ou satellites occupe la communauté de télédétection depuis longtemps, par le biais de processus de classification supervisés, tels que les SVMs, ou les forêts aléatoires pour, entre autres, la vitesse d’exécution de ces derniers. Mais les résultats obtenus par ces méthodes n'ont pas encore permis une réelle automatisation, notamment en adéquation avec des spécifications existantes (erreurs encore trop importantes). En parallèle de ces algorithmes depuis longtemps utilisés, des méthodes d’apprentissage automatique d’un genre nouveau, bien que reposant sur des concepts remontant aux années 1950, émergent depuis une décennie. L’apprentissage profond a fait ses preuves dans de nombreux domaines depuis le traitement naturel du langage, à la reconnaissance d’objets dans des images. Cet essor récent est la conséquence de la disponibilité de grandes bases de données d'apprentissage, ainsi que la croissance globale des puissances de calcul. Représentants principaux de cette famille d’apprentissage, les réseaux de neurones profonds ont un impact au quotidien. Que ce soit au niveau académique en terme de recherche, au niveau sociétal, au travers des smartphones par exemple, ou même au niveau politique, avec les questions déontologiques que cela peut poser en terme de confidentialité des données (RGPD) et de protection des libertés individuelles, l'apprentissage profond est au cœur de technologies utilisées par la plupart des gens à leur insu. En effet, pour afficher de telles performances dans tant de domaines, ces algorithmes nécessitent de gros volumes de données. Les bases de données géographiques de l’IGN sont donc une opportunité dans notre cas, permettant d’exploiter au mieux les images très haute résolution monoscopiques acquises par les satellites SPOT 6 et 7 en les classifiant automatiquement par réseaux de neurones profonds appris sur ces mêmes bases de données. C’est cette approche que nous proposons dans ces travaux de thèse, avec une volonté d’étudier cette problématique tout en se plaçant dans un cadre plus large à visée opérationnelle. Nous étudions les possibilités de transfert d’apprentissage qui offre beaucoup d’avantages en matière de charges de calcul et de jeu d’apprentissage. Enfin, l’exploitation jointe d’images aériennes et de réseaux de neurones profonds est étudiée, avec un accent mis sur la préparation des données d’apprentissage issues des bases de données géographiques de l’IGN qui présentent certains inconvénients

International audience; The dynamical properties of so-called spatial and complex networks are often overlooked in graph theory and network science in general. Container shipping provides a rare example of a global transport network that went through tremendous technological and geographic changes in the last decades or so. This chapter proposes for the first time an empirical analysis of no less than 40 years of inter-port vessel movement data (1977-2016) to describe the evolving properties of the global container shipping network. Main results confirm a number of stylized facts such as the growing size, connectivity, and centralization of this network due to several factors such as economies of scale in liner shipping and the rationalization of related maritime services, the emergence of hub ports, etc. We also provide a new cartography of how had the global container shipping network been geographically distributed over time, thereby highlighting major shifts in terms of port hierarchies and main corridors. We believe that this chapter will contribute to a better understanding of the complex linkages between network structure, technological change, and spatial change, opening the way for new research paths on maritime transport research and network science in general when focusing on evolutionary dynamics.

Following the first initiatives taken by the International Programme Committee of the XXIIIrd ISPRS Congress in Prague (Czech Republic) in 2016, modifications of the reviewing process of ISPRS events were further considered during the years 2017 and 2018. This evolution first targets to better fit such a process to the currents requirements and expectations of the ISPRS community. Secondly, it aims to provide unified guidelines for the different steps of the process. Under the aegis of the 2020 Congress Director and ISAC (International Science Advisory Committee) chair, several discussions were held in-between September 2017 and June 2018 with ISAC members, Technical Commission Presidents (TCP), council members, 2016 and 2020 Congress Programme Chairs. This document serves as a unique transparent basis that applies for all kinds of ISPRS events (from Congress and Geospatial Week to smaller workshops), and all categories of people that are bound to be involved in the evaluation process of scientific contributions (authors, reviewers, TCPs, … ). It also specifies the evaluation criteria for the works submitted to ISPRS events, both for full papers and abstracts. Subsequently, it helps authors to improve the content and shape of their contributions. Eventually, this paper is targeted to help new chairs to smoothly prepare their future event. The following guidelines were first adopted for the 2018 Technical Commission Symposia.

International audience; We propose a new method for the reconstruction of simplicial complexes (combining points, edges and triangles) from 3D point clouds from Mobile Laser Scanning (MLS). Our main goal is to produce a reconstruction of a scene that is adapted to the local geometry of objects. Our method uses the inherent topology of the MLS sensor to define a spatial adjacency relationship between points. We then investigate each possible connexion between adjacent points and filter them by searching collinear structures in the scene, or structures perpendicular to the laser beams. Next, we create triangles for each triplet of self-connected edges. Last, we improve this method with a regularization based on the co-planarity of triangles and collinearity of remaining edges. We compare our results to a naive simplicial complexes reconstruction based on edge length.; Nous présentons une nouvelle méthode pour la reconstruction de complexes simpliciaux (ensembles de points, segments et triangles) à partir de nuages de points 3D obtenus par LiDAR mobile, à balayage plan. Notre méthode utilise la topo-logie inhérente au capteur LiDAR pour définir une relation spatiale entre les points. Pour cela, nous examinons chaque connexion possible entre points, pondérée en fonction de sa distance au capteur, et les filtrons en privilégiant les structures collinéaires, ou perpendiculaires aux impulsions du laser. Ensuite, nous créons et filtrons des triangles pour chaque triplet de segments connectés entre eux, en fonction de leur coplanarité locale. Nous comparons nos résultats à une reconstruction non pondérée d'un complexe simplicial. Abstract We propose a new method for the reconstruction of simplicial complexes (combining points, edges and triangles) from 3D point clouds from plane sweep Mobile Laser Scanning (MLS). Our method uses the inherent topology of the MLS sensor to define a spatial adjacency relationship between points. We then investigate each possible connexion between adjacent points, weighted according to its distance to the sensor, and filter them by searching collinear structures in the scene, or structures perpendicular to the laser beams. Next, we create and filter triangles for each triplet of self-connected edges and according to their local planarity. We compare our results to an unweighted simplicial complex reconstruction.

version étendue publiée : hal-02155541; International audience; Automatic map generalization requires the use of computationally intensive processes often unable to deal with large datasets. Distributing the generalization process is the only way to make them scalable and usable in practice. But map generalization is a highly contextual process, and the surroundings of a generalized map feature needs to be known to generalize the feature, which is a problem as distribution might partition the dataset and parallelize the processing of each part. This paper proposes experiments to evaluate the past propositions to distribute map generalization, and to identify the main remaining issues. The past propositions to distribute map generalization are first discussed, and then the experiment hypotheses and apparatus are described. The experiments confirmed that regular partitioning was the quickest strategy, but also the less effective in taking context into account. The geographical partitioning, though less effective for now, is quite promising regarding the quality of the results as it better integrates the geographical context.

Background T‐type calcium channels have been shown to play an important role in the initiation and maintenance of neuropathic pain and represent a promising therapeutic target for new analgesic treatments. Ethosuximide (ETX), an anticonvulsant and a T‐type channel blocker has shown analgesic effect in several chronic pain models but has not yet been evaluated in patients with neuropathic pain. Methods This proof‐of‐concept, multicentre, double‐blind, controlled and randomized trial compared the efficacy and safety of ETX (given as add‐on therapy) to an inactive control (IC) in 114 patients with non‐diabetic peripheral neuropathic pain. After a 7‐day run‐in period, eligible patients aged over 18 years were randomly assigned (1:1) to ETX or IC for 6 weeks. The primary outcome was the difference between groups in the pain intensity (% of change from the baseline to end of treatment) assessed in the intention‐to‐treat population. This study is registered with EudraCT (2013‐004801‐26) and ClinicalTrials.gov (NCT02100046). Results The study was stopped during the interim analysis due to the high number of adverse events in the active treatment group. ETX failed to reduce total pain and showed a poor tolerance in comparison to IC. In the per‐protocol analysis, ETX significantly reduced pain intensity by 15.6% (95% CI −25.8; −5.4) from baseline compared to IC (−7.8%, 95% CI −14.3; −1.3; p = 0.033), but this result must be interpreted with caution because of a small subgroup of patients. Conclusion Ethosuximide did not reduce the severity of neuropathic pain and induces, at the doses used, many adverse events. Significance This article shows that ETX is not effective to treat neuropathic pain. Nevertheless, per‐protocol analysis suggests a possible analgesic effect of ETX. Thus, our work adds significant knowledge to preclinical and clinical data on the benefits of T‐type calcium channel inhibition for the treatment of neuropathic pain.

While annual crop rotations play a crucial role for agricultural optimization, they have been largely ignored for automated crop type mapping. In this paper, we take advantage of the increasing quantity of annotated satellite data to propose the first deep learning approach modeling simultaneously the inter- and intra-annual agricultural dynamics of parcel classification. Along with simple training adjustments, our model provides an improvement of over 6.3 mIoU points over the current state-of-the-art of crop classification. Furthermore, we release the first large-scale multi-year agricultural dataset with over 300,000 annotated parcels., Comment: Published in Remote Sensing

International audience; This article presents a coupled system consisting of a single-frequency GPS receiver and a light photogrammetric quality camera embedded in an Unmanned Aerial Vehicle (UAV). The aim is to produce high quality data that can be used in metrology applications. The issue of Integrated Sensor Orientation (ISO) of camera poses using only GPS measurements is presented and discussed. The accuracy reached by our system based on sensors developed at the French Mapping Agency (IGN) Opto-Electronics, Instrumentation and Metrology Laboratory (LOEMI) is qualified. These sensors are specially designed for close-range aerial image acquisition with a UAV. Lever-arm calibration and time synchronization are explained and performed to reach maximum accuracy. All processing steps are detailed from data acquisition to quality control of final products. We show that an accuracy of a few centimeters can be reached with this system which uses low-cost UAV and GPS module coupled with the IGN-LOEMI home-made camera.

In this paper, we focus on the automatic interpretation of 3D point cloud data in terms of associating a class label to each 3D point. While much effort has recently been spent on this research topic, little attention has been paid to the influencing factors that affect the quality of the derived classification results. For this reason, we investigate fundamental influencing factors making geometric features more or less relevant with respect to the classification task. We present a framework which consists of five components addressing point sampling, neighborhood recovery, feature extraction, classification and feature relevance assessment. To analyze the impact of the main influencing factors which are represented by the given point sampling and the selected neighborhood type, we present the results derived with different configurations of our framework for a commonly used benchmark dataset for which a reference labeling with respect to three structural classes (linear structures, planar structures and volumetric structures) as well as a reference labeling with respect to five semantic classes (Wire, Pole/Trunk, Façade, Ground and Vegetation) is available.