Search

At present, with the advance of satellite image processing technology, remote sensing images are becoming more widely used in real scenes. However, due to the limitations of current remote sensing imaging technology and the influence of the external environment, the resolution of remote sensing images often struggles to meet application requirements. In order to obtain high-resolution remote sensing images, image super-resolution methods are gradually being applied to the recovery and reconstruction of remote sensing images. The use of image super-resolution methods can overcome the current limitations of remote sensing image acquisition systems and acquisition environments, solving the problems of poor-quality remote sensing images, blurred regions of interest, and the requirement for high-efficiency image reconstruction, a research topic that is of significant relevance to image processing. In recent years, there has been tremendous progress made in image super-resolution methods, driven by the continuous development of deep learning algorithms. In this paper, we provide a comprehensive overview and analysis of deep-learning-based image super-resolution methods. Specifically, we first introduce the research background and details of image super-resolution techniques. Second, we present some important works on remote sensing image super-resolution, such as training and testing datasets, image quality and model performance evaluation methods, model design principles, related applications, etc. Finally, we point out some existing problems and future directions in the field of remote sensing image super-resolution. [ABSTRACT FROM AUTHOR]

The occurrence of forest fires has increased significantly in recent years across the planet. Events of this nature have resulted in the leveraging of new automated methodologies to identify and map burned areas. In this paper, we introduce a unified data-driven framework capable of mapping areas damaged by fire by integrating time series of remotely sensed multispectral images, statistical modeling, and unsupervised classification. We collect and analyze multiple remote-sensing images acquired by the Landsat-8, Sentinel-2, and Terra satellites between August–October 2020, validating our proposal with three case studies in Brazil and Bolivia whose affected regions have suffered from recurrent forest fires. Besides providing less noisy mappings, our methodology outperforms other evaluated methods in terms of average scores of 90%, 0.71, and 0.65 for overall accuracy, F1-score, and kappa coefficient, respectively. The proposed method provides spatial-adherence mappings of the burned areas whose segments match the estimates reported by the MODIS Burn Area product. [ABSTRACT FROM AUTHOR]

To date, no method utilizing satellite imagery exists for detailing the locations and functions of critical infrastructure across the United States, making response to natural disasters and other events challenging due to complex infrastructural interdependencies. This paper presents a repeatable, transferable, and explainable method for critical infrastructure analysis and implementation of a robust model for critical infrastructure detection in satellite imagery. This model consists of a DenseNet-161 convolutional neural network, pretrained with the ImageNet database. The model was provided additional training with a custom dataset, containing nine infrastructure classes. The resultant analysis achieved an overall accuracy of 90%, with the highest accuracy for airports (97%), hydroelectric dams (96%), solar farms (94%), substations (91%), potable water tanks (93%), and hospitals (93%). Critical infrastructure types with relatively low accuracy are likely influenced by data commonality between similar infrastructure components for petroleum terminals (86%), water treatment plants (78%), and natural gas generation (78%). Local interpretable model-agnostic explanations (LIME) was integrated into the overall modeling pipeline to establish trust for users in critical infrastructure applications. The results demonstrate the effectiveness of a convolutional neural network approach for critical infrastructure identification, with higher than 90% accuracy in identifying six of the critical infrastructure facility types. [ABSTRACT FROM AUTHOR]

Forest fires are considered to be an important part of numerous terrestrial ecosystems and vegetation types, being also a significant factor of ecosystem disruption. In this sense, fires play an important role in the structure and function of the ecosystems. Biomes are characterized by a specific type of fire regime, which is a synergy of the climate conditions and the characteristics of the vegetation types dominating each biome. The assessment of burned areas and the identification of the fire regimes can be implemented with freely available low- to high-resolution satellite data as those of Landsat and Sentinel-2. Moreover, the biomes are characterized by the phenology, a useful component for vegetation monitoring, especially when time series of satellite images are used. Both the identification of fire regime by reconstructing the fire history and the monitoring of the post-fire evolution of burned areas were studied with remote sensing methods. Specifically, the present paper is a pilot study implemented in a Mediterranean biome, aimed at establishing the methodological framework to (i) define fire regimes, (ii) characterize the phenological pattern of the vegetation (pre-fire situation) of the fire-affected areas, and (iii) compare the phenology of the recovered fire-affected areas with the corresponding one of the pre-fire situation. At the global level, based on MODIS fire perimeters, we found that fires are occurring at 70% in the tropical and subtropical grasslands, savannas, and shrublands, followed by fires at tropical and subtropical moist broadleaf forests by 7% and by fires at deserts and xeric shrublands by 6.5%. [ABSTRACT FROM AUTHOR]

Background: Road network data are crucial in various applications, such as emergency response, urban planning, and transportation management. The recent application of deep neural networks has significantly boosted the efficiency and accuracy of road network extraction based on remote sensing data. However, most existing methods for road extraction were designed at local or regional scales. Automatic extraction of large-scale road datasets from satellite images remains challenging due to the complex background around the roads, especially the complicated land cover types. To tackle this issue, this paper proposes a land cover background-adaptive framework for large-scale road extraction. Method: A large number of sample image blocks (6820) are selected from six different countries of a wide region as the dataset. OpenStreetMap (OSM) is automatically converted to the ground truth of networks, and Esri 2020 Land Cover Dataset is taken as the background land cover information. A fuzzy C-means clustering algorithm is first applied to cluster the sample images according to the proportion of certain land use types that obviously negatively affect road extraction performance. Then, the specific model is trained on the images clustered as abundant with that certain land use type, while a general model is trained based on the rest of the images. Finally, the road extraction results obtained by those general and specific modes are combined. Results: The dataset selection and algorithm implementation were conducted on the cloud-based geoinformation platform Google Earth Engine (GEE) and Google Colaboratory. Experimental results showed that the proposed framework achieved stronger adaptivity on large-scale road extraction in both visual and statistical analysis. The C-means clustering algorithm applied in this study outperformed other hard clustering algorithms. Significance: The promising potential of the proposed background-adaptive network was demonstrated in the automatic extraction of large-scale road networks from satellite images as well as other object detection tasks. This search demonstrated a new paradigm for the study of large-scale remote sensing applications based on deep neural networks. [ABSTRACT FROM AUTHOR]

The spatial distribution of remote-sensing scene images is highly complex in character, so how to extract local key semantic information and discriminative features is the key to making it possible to classify accurately. However, most of the existing convolutional neural network (CNN) models tend to have global feature representations and lose the shallow features. In addition, when the network is too deep, gradient disappearance and overfitting tend to occur. To solve these problems, a lightweight, multi-instance CNN model for remote sensing scene classification is proposed in this paper: MILRDA. In the instance extraction and classifier part, more discriminative features are extracted by the constructed residual dense attention block (RDAB) while retaining shallow features. Then, the extracted features are transformed into instance-level vectors and the local information associated with bag-level labels is highlighted by the proposed channel-attention-based multi-instance pooling, while suppressing the weights of useless objects or backgrounds. Finally, the network is constrained by the cross-entropy loss function to output the final prediction results. The experimental results on four public datasets show that our proposed method can achieve comparable results to other state-of-the-art methods. Moreover, the visualization of feature maps shows that MILRDA can find more effective features. [ABSTRACT FROM AUTHOR]

Atmospheric correction (AC) is the basis for quantitative water remote sensing, and adjacency effects form an important part of AC for medium- and high-spatial-resolution optical satellite images. The 6S radiative transfer model is widely used, but its background reflectance function does not take the radiance changes at water–land boundaries into account. If the observed land possesses bright features, the radiance of the adjacent water will be affected, leading to deviations in the AC results and increasing the uncertainty of water depth-based optical quantitative remote sensing. In this paper, we propose a model named WL-AE (a correction model for water–land boundary adjacency effects), which is based on the obvious radiance differences at water–land boundaries. This model overcomes the problem by which the background reflectance calculation is not terminated due to the highlighting pixel. We consider the influences of different R n s (neighborhood space) on the target pixel. The effective calculation of the equivalent background reflectance of the target pixel is realized, and the influence of the land area anomaly highlighting the pixel on the adjacent water is avoided. The results show that WL-AE can effectively improve the entropy and contrast of the input image and that the water–land boundary is greatly affected by adjacency effects, especially in the green and near-infrared bands, where the M r c (mean rate of change) are as high as 14.2% and 20.1%, respectively. In the visible wavelength, the S d of R r c (the relative rate of change) is positively correlated with R n s , and the S d reaches 16.9%. Although the adjacency effect is affected by ground object types, its influence area remains within 3 km offshore. Based on the WL-AE and 6S results, the comparative test regarding bathymetric inversion shows that the influence is significant in the 0–5 m depth section. In Penang, the MRE of the 0–4 m inversion results is 31.4%, which is 10.5% lower than that of the 6S model. [ABSTRACT FROM AUTHOR]

The number and area of alpine lakes in Tien Shan (TS) are rapidly growing in response to a warming climate and retreating glaciers. This paper presents a comparative analysis of lake classification and changes by dividing alpine lakes (within a 10 km buffer of the glacier margins) into four types (supraglacial lakes, proglacial lakes, extraglacial lakes and non-glacial lakes), and subsequently determining the driving forces of change across the TS region from 1990 to 2015. The analysis utilized multiple satellite images and climatic data from gridded data sets and meteorological station observations. The results indicate that the total number and area of glacial lakes continuously increased during the study period, whereas non-glacial lakes intermittently expanded. Specifically, the total number and area of all glacial lakes (supraglacial lakes, proglacial lakes and extraglacial lakes) increased by 45.45% and 27.08%, respectively. Non-glacial lakes, in contrast, increased in quantity and area by 23.92% and 19.01%, respectively. Alpine lakes are closer to glaciers at high altitudes; in fact, some (e.g., proglacial lakes) are connected to glacier termini, and these show the highest expansion speed during the study period. The area of proglacial lakes expanded by 60.32%. Extraglacial lakes expanded by 21.06%. Supraglacial lakes, in marked contrast to the other types, decreased in area by 3.74%. Widespread rises in temperature and glacier wastage were the primary cause of the steady expansion of glacial lakes, particularly those linked to small- and medium-sized glaciers distributed in the Eastern TS where glacial lakes have rapidly increased. Both proglacial and extraglacial lakes expanded by 6.47%/a and 2%/a, respectively, from 1990 to 2015. While these proglacial and extraglacial lakes are located in largely glacierized areas, lakes in the Central TS exhibited the slowest expansion, increasing in area by 1.44%/a and 0.74%/a, respectively. Alterations in non-glacial lake areas were driven by changes in precipitation and varied spatially over the region. This study has substantial implications for the state of water resources under the complex regional changes in climate in the TS and can be used to develop useful water-resource management and planning strategies throughout Central Asia. [ABSTRACT FROM AUTHOR]

With the increasing number of reservoirs on the Nile River Basin, it has become important to understand the reservoir operations in the basin for coordinated water management among the various countries. With the lack of a proper framework for data sharing amongst the Nile basin countries, satellite remote sensing provides a simple transparent way to continuously monitor the changes taking place in reservoirs in all regions of the Nile River Basin. This paper presents a comparison between Sentinel-1- and Sentinel-2-derived reservoir water levels and the altimetry-based water level from G-REALM (Global Reservoirs and Lakes Monitor) for three major reservoirs downstream of the Millennium Reservoir impounded by the Grand Ethiopian Renaissance Dam (GERD) on the Nile River for the period of 2014–2021. Water surface extents were derived from Sentinel-1 using dynamic thresholds and from Sentinel-2 with the use of the NDWI (Normalized Difference Water Index). The water levels were estimated using a DEM-based contour matching technique. For Roseires Reservoir, the water levels from Sentinel agreed well with those from G-REALM (RMSE = 0.92 m; R2 = 0.82). For Lake Nasser, the water levels also agreed well (RMSE = 0.72 m; R2 = 0.85). For Lake Merowe, there was a significant mismatch in the derived water levels, mostly due to a lack of sufficient data from both sources. Overall, satellite imagery from Sentinel provides a very good alternative to altimetry-based water levels for the Nile River Basin. [ABSTRACT FROM AUTHOR]

Domain adaptation for classification has achieved significant progress in natural images but not in remote-sensing images due to huge differences in data-imaging mechanisms between different modalities and inconsistencies in class labels among existing datasets. More importantly, the lack of cross-domain benchmark datasets has become a major obstacle to the development of scene classification in multimodal remote-sensing images. In this paper, we present a cross-domain dataset of multimodal remote-sensing scene classification (MRSSC). The proposed MRSSC dataset contains 26,710 images of 7 typical scene categories with 4 distinct domains that are collected from Tiangong-2, a Chinese manned spacecraft. Based on this dataset, we evaluate several representative domain adaptation algorithms on three cross-domain tasks to build baselines for future research. The results demonstrate that the domain adaptation algorithm can reduce the differences in data distribution between different domains and improve the accuracy of the three tasks to varying degrees. Furthermore, MRSSC also achieved fairly results in three applications: cross-domain data annotation, weakly supervised object detection and data retrieval. This dataset is believed to stimulate innovative research ideas and methods in remote-sensing cross-domain scene classification and remote-sensing intelligent interpretation. [ABSTRACT FROM AUTHOR]

Remote sensing image scene classification takes image blocks as classification units and predicts their semantic descriptors. Because it is difficult to obtain enough labeled samples for all classes of remote sensing image scenes, zero-shot classification methods which can recognize image scenes that are not seen in the training stage are of great significance. By projecting the image visual features and the class semantic features into the latent space and ensuring their alignment, the variational autoencoder (VAE) generative model has been applied to address remote-sensing image scene classification under a zero-shot setting. However, the VAE model takes the element-wise square error as the reconstruction loss, which may not be suitable for measuring the reconstruction quality of the visual and semantic features. Therefore, this paper proposes to augment the VAE models with the generative adversarial network (GAN) to make use of the GAN's discriminator in order to learn a suitable reconstruction quality metric for VAE. To promote feature alignment in the latent space, we have also proposed cross-modal feature-matching loss to make sure that the visual features of one class are aligned with the semantic features of the class and not those of other classes. Based on a public dataset, our experiments have shown the effects of the proposed improvements. Moreover, taking the ResNet models of ResNet18, extracting 512-dimensional visual features, and ResNet50 and ResNet101, both extracting 2048-dimensional visual features for testing, the impact of the different visual feature extractors has also been investigated. The experimental results show that better performance is achieved by ResNet18. This indicates that more layers of the extractors and larger dimensions of the extracted features may not contribute to the image scene classification under a zero-shot setting. [ABSTRACT FROM AUTHOR]

Within water resources management, surface water area (SWA) variation plays a vital role in hydrological processes as well as in agriculture, environmental ecosystems, and ecological processes. The monitoring of long-term spatiotemporal SWA changes is even more critical within highly populated regions that have an arid or semi-arid climate, such as Iran. This paper examined variations in SWA in Iran from 1990 to 2021 using about 18,000 Landsat 5, 7, and 8 satellite images through the Google Earth Engine (GEE) cloud processing platform. To this end, the performance of twelve water mapping rules (WMRs) within remotely-sensed imagery was also evaluated. Our findings revealed that (1) methods which provide a higher separation (derived from transformed divergence (TD) and Jefferies–Matusita (JM) distances) between the two target classes (water and non-water) result in higher classification accuracy (overall accuracy (OA) and user accuracy (UA) of each class). (2) Near-infrared (NIR)-based WMRs are more accurate than short-wave infrared (SWIR)-based methods for arid regions. (3) The SWA in Iran has an overall downward trend (observed by linear regression (LR) and sequential Mann–Kendall (SQMK) tests). (4) Of the five major water basins, only the Persian Gulf Basin had an upward trend. (5) While temperature has trended upward, the precipitation and normalized difference vegetation index (NDVI), a measure of the country's greenness, have experienced a downward trend. (6) Precipitation showed the highest correlation with changes in SWA (r = 0.69). (7) Long-term changes in SWA were highly correlated (r = 0.98) with variations in the JRC world water map. [ABSTRACT FROM AUTHOR]

The high-precision geometric positioning of optical remote sensing satellites is the prerequisite to determine the application capability of satellite image products. Its positioning accuracy is related to the observation accuracy of each link in the imaging process, including satellite attitude, orbit measurement accuracy, time synchronization accuracy, camera measurement accuracy, and so on. Untimely and inaccurate on-orbit calibration will lead to great geometric positioning errors. To optimize the positioning accuracy of satellite images with the rational function model (RFM) under low positioning accuracy, our paper proposes an improved geometric quality model based on the reorientation of internal and external orientation elements in the RFM model of remote sensing images. By establishing the rational function positioning model, the external orientation model, and the internal orientation model, the original image can be reorientated. Then, we use the improved model to generate uniformly distributed virtual ground control points. By analyzing and verifying the relationship between each rational polynomial coefficient (RPC) and its influence on geometric positioning accuracy, we propose an RPC coefficients optimization method based on image offset correction and positioning dominant coefficients. Finally, we use the small satellite "MN200Sar-1" with low geometric accuracy for experimental verification. The results show that the model can effectively eliminate the errors of internal and external elements in the on-orbit calibration, and the positioning accuracy is improved from one hundred pixels to one pixel. At the same time, the rational polynomial dominant coefficient optimization method can improve geometric positioning accuracy without introducing additional compensation parameters. [ABSTRACT FROM AUTHOR]

In recent years, the conversion of saline-alkali land to rice fields has become the most dominant land use change feature in western Jilin, leading to significant surface greening. Saline–alkali land and paddy fields have distinct surface biophysical properties; however, there is a lack of systematic assessment of the moderating effect of planting rice on saline–alkali land on regional climate by changing surface properties. In this paper, multiscale data on the surface temperature of saline–alkali land and paddy fields were obtained using 1 km MODIS product, 30 m Landsat 8 satellite imagery and centimeter-scale UAV imagery in Da'an City, western Jilin as the study area, and the various characteristics of the surface temperature of saline-alkali land and paddy fields in different months of the year and at different times of the day were analyzed. Furthermore, the effect of rice cultivation in saline–alkali land on the local surface temperature was assessed using a space-for-time approach. The results based on satellite observations including both MODIS and Landsat showed that the surface temperature of saline–alkali land was significantly higher than that of paddy fields during the crop growing season, especially in July and August. The high temporal resolution MODIS LST data also indicated the paddy fields cool the daytime surface temperature, while warming the nighttime surface temperature, which was in contrast for saline–alkali land during the growing season. High-resolution UAV observations in July confirmed that the cooling effect of paddy fields was most significant at the middle of day. From the biophysical perspective, the reclamation of saline–alkali land into paddy fields leads to an increase in leaf area index, followed by a significant increase in evapotranspiration. Meanwhile, rice cultivation in saline–alkali land reduces surface albedo and increases surface net radiation. The trade-off relationship between the two determines the seasonal difference in the surface temperature response of saline–alkali land for rice cultivation. At the same time, the daily cycle of crop evapotranspiration and the thermal insulation effect of paddy fields at night are the main reasons for the intraday difference in surface temperature between saline–alkali land and paddy field. Based on the multiscale assessment of the impact of rice cultivation in saline-alkali land on surface temperature, this study provides a scientific basis for predicting future regional climate change and comprehensively understanding the ecological and environmental benefits of saline–alkali land development. [ABSTRACT FROM AUTHOR]

This paper presents the development and update of a multi-scale yield prediction model for processing tomatoes. The study was carried out under the EU-funded programme "Support to Development of a Rural Business Information System (RBIS)", and the performance of the updated crop-specific yield prediction models and their generated predictions at regional and national levels are presented. The model was built using Sentinel-2 satellite imagery to obtain cumulative values of six (6) selected vegetation indices (VIs). The data were collected on five (5) different dates for processing tomato fields in the Khachmaz region of Azerbaijan during summer 2021 (June to August) at 10- to 13-day intervals. In addition, a targeted field sampling campaign was conducted on selected Khachmaz pilot fields towards the end of the growing season to assess the potential of Sentinel-2 data to determine yield variability in tomato fields. Finally, actual recorded yields were collected at the field level to build the yield prediction regression model and evaluate its performance at different spatial scales, ranging from single field to national level, as well as under different data availability scenarios (number of consecutive Sentinel-2 images used). The results showed a high degree of correlation between all implemented VIs and processing tomato yield, with a coefficient of determination of up to 0.89 for the NDVI, providing valuable information for future estimates of tomato production across multiple spatial scales. The developed prediction model could also be used in the agri-food sector for national yield estimates to support policy and regulatory decisions at the national level. [ABSTRACT FROM AUTHOR]

Individual tree counting (ITC) is a popular topic in the remote sensing application field. The number and planting density of trees are significant for estimating the yield and for futher planing, etc. Although existing studies have already achieved great performance on tree detection with satellite imagery, the quality is often negatively affected by clouds and heavy fog, which limits the application of high-frequency inventory. Nowadays, with ultra high spatial resolution and convenient usage, Unmanned Aerial Vehicles (UAVs) have become promising tools for obtaining statistics from plantations. However, for large scale areas, a UAV cannot capture the whole region of interest in one photo session. In this paper, a real-time orthophoto mosaicing-based tree counting framework is proposed to detect trees using sequential aerial images, which is very effective for fast detection of large areas. Firstly, to guarantee the speed and accuracy, a multi-planar assumption constrained graph optimization algorithm is proposed to estimate the camera pose and generate orthophoto mosaicing simultaneously. Secondly, to avoid time-consuming box or mask annotations, a point supervised method is designed for tree counting task, which greatly speeds up the entire workflow. We demonstrate the effectiveness of our method by performing extensive experiments on oil-palm and acacia trees. To avoid the delay between data acquisition and processing, the proposed framework algorithm is embedded into the UAV for completing tree counting tasks, which also reduces the quantity of data transmission from the UAV system to the ground station. We evaluate the proposed pipeline using sequential UAV images captured in Indonesia. The proposed pipeline achieves an F1-score of 98.2% for acacia tree detection and 96.3% for oil-palm tree detection with online orthophoto mosaicing generation. [ABSTRACT FROM AUTHOR]

A miniaturized video satellite can observe the ground targets by recording real-time video clips in staring control mode and therefore obtains a unique advantage over traditional remote sensing techniques. To further extend the application of a video satellite, a strategy for simultaneously observing a group of ground targets is to be developed. To cope with the impacts of an uncalibrated camera on the pointing accuracy which can lead to the failure of a multi-target observation task, an adaptive attitude control method is to be exploited. Hence, to observe multiple ground targets using an onboard uncalibrated camera, this paper proposes an image-based adaptive staring attitude controller. First, a target-selection strategy is proposed to realize a more balanced staring observation of the target group. Second, an updating law is proposed to estimate the camera parameters according to the projection equations. At last, an adaptive staring controller based on the estimated parameters is formulated, so that the center of mass of the ground targets on the image can be controlled towards its desired location, which is normally the image center. The stability of the proposed staring controller is proved using Barbalat's Lemma. The simulation results show that even though the camera parameters are uncertain, the adaptive control method effectively achieves the staring observation for multiple ground targets by keeping their midpoint at the image center. [ABSTRACT FROM AUTHOR]

Aircraft-Induced Clouds (AICs), colloquially called contrails, form from the emission of soot from jet engines during cruise flight in favorable atmospheric conditions. AICs absorb, scatter, and reflect shortwave and longwave radiation. This radiative transfer has a cooling effect during the day; however, the night experiences an overwhelming warming effect, which leads to an overall warming effect on Earth, contributing to anthropogenically propelled climate change. Reducing AICs significantly mitigates aviation's contribution to climate change by reducing the disruption in Earth's radiation budget. Researchers have proposed AIC Abatement Programs (AAPs) to increase cruise flight levels without additional fuel burn. In order to effectively implement AAPs, it is crucial to be able to accurately identify AICs from publicly available aerial and satellite imagery. This study aims at the identification of AICs from hyperspectral imagery to help the effective implementation of an AAP and to mitigate climate change. This paper describes a method for the hyperspectral analysis of aerial images in order to accurately identify AICs through a case study based in West Virginia. The results show that both the Adaptive Coherence Estimator and the Matched Filter algorithms based on unique in-scene spectra were successful in the isolation of the AICs from other cloud types and the background. It is found that AICs can be identified with 84% confidence in this case study. The method, a case study, and future works are provided. [ABSTRACT FROM AUTHOR]

Change detection between images of pre-flood and flooding periods is a critical process for flood mapping using satellite images. Flood mapping from SAR images is based on backscattering coefficient differences. The change rules of the backscattering coefficient with different flooding depths of ground objects are essential prior knowledge for flood mapping, while their absence greatly limits the precision. Therefore, minimizing the backscattering coefficient differences caused by non-flood factors is of great significance for improving the accuracy of flood mapping. In this paper, non-flood factor influences, i.e., monthly variations of ground objects and polarization and satellite orbits, on the backscattering coefficient are studied with multi-temporal Sentinel-1 images for five ground objects in Kouzi Village, Shouguang City, Shandong Province, China. Sentinel-1 images in different rainfalls are used to study the variation of the backscattering coefficient with flooding depths. Since it is difficult to measure the flooding depth of historical rainfall events, a hydrological analysis based on the Geographic Information System (GIS) and Remote Sensing (RS) is used to estimate the flooding depth. The results showed that the monthly variations of the maximum backscattering coefficients of farmland and construction and the backscattering coefficient differences caused by the satellite orbit were larger than the minimum backscattering coefficient differences caused by inundation. The flood extraction rules of five objects based on Sentinel-1 were obtained and analyzed, which improved flood extraction knowledge from qualitative to semi-quantitative analysis. [ABSTRACT FROM AUTHOR]

The technological development of the remote sensing domain led to the acquisition of satellite image time series (SITS) for Earth Observation (EO) by a variety of sensors. The variability in terms of the characteristics of the satellite sensors requires the existence of algorithms that allow the integration of multiple modalities and the identification of anomalous spatio-temporal evolutions caused by natural hazards. The unsupervised analysis of multimodal SITS proposed in this paper follows a two-step methodology: (i) inter-modality translation and (ii) the identification of anomalies in a change-detection framework. Inter-modality translation is achieved by means of a Generative Adversarial Network (GAN) architecture, whereas, for the identification of anomalies caused by natural hazards, we adapt the task to a similarity search in SITS. In this regard, we provide an extension of the matrix profile concept, which represents an answer to identifying differences and to discovering novelties in time series. Furthermore, the proposed inter-modality translation allows the usage of standard unsupervised clustering approaches (e.g., K-means using the Dynamic Time Warping measure) for mono-modal SITS analysis. The effectiveness of the proposed methodology is shown in two use-case scenarios, namely flooding and landslide events, for which a joint acquisition of Sentinel-1 and Sentinel-2 images is performed. [ABSTRACT FROM AUTHOR]

Vehicle detection is an important but challenging problem in Earth observation due to the intricately small sizes and varied appearances of the objects of interest. In this paper, we use these issues to our advantage by considering them results of latent image augmentation. In particular, we propose using supervised contrastive loss in combination with a mutual guidance matching process to helps learn stronger object representations and tackles the misalignment of localization and classification in object detection. Extensive experiments are performed to understand the combination of the two strategies and show the benefits for vehicle detection on aerial and satellite images, achieving performance on par with state-of-the-art methods designed for small and very small object detection. As the proposed method is domain-agnostic, it might also be used for visual representation learning in generic computer vision problems. [ABSTRACT FROM AUTHOR]

Change detection is an important task in remote-sensing image analysis. With the widespread development of deep learning in change detection, most of the current methods improve detection performance by making the network deeper and wider, but ignore the inference time and computational costs of the network. Therefore, this paper proposes a lightweight change-detection network called Shuffle-CDNet. It accepts the six-channel image that concatenates the bitemporal images by channel as the input, and it adopts the backbone network with channel shuffle operation and depthwise separable convolution layers. The classifier uses a lightweight atrous spatial pyramid pooling (Light-ASPP) module to reduce computational costs. The edge-information feature extracted by a lightweight branch is integrated with the shallow and deep features extracted by the backbone network, and the spatial and channel attention mechanisms are introduced to enhance the expression of features. At the same time, logit knowledge distillation and data augmentation techniques are used in the training phase to improve detection performance. Experimental results showed that the proposed method achieves a better balance in computational efficiency and detection performance compared with other advanced methods. [ABSTRACT FROM AUTHOR]

In supervised deep learning, learning good representations for remote-sensing images (RSI) relies on manual annotations. However, in the area of remote sensing, it is hard to obtain huge amounts of labeled data. Recently, self-supervised learning shows its outstanding capability to learn representations of images, especially the methods of instance discrimination. Comparing methods of instance discrimination, clustering-based methods not only view the transformations of the same image as "positive" samples but also similar images. In this paper, we propose a new clustering-based method for representation learning. We first introduce a quantity to measure representations' discriminativeness and from which we show that even distribution requires the most discriminative representations. This provides a theoretical insight into why evenly distributing the images works well. We notice that only the even distributions that preserve representations' neighborhood relations are desirable. Therefore, we develop an algorithm that translates the outputs of a neural network to achieve the goal of evenly distributing the samples while preserving outputs' neighborhood relations. Extensive experiments have demonstrated that our method can learn representations that are as good as or better than the state of the art approaches, and that our method performs computationally efficiently and robustly on various RSI datasets. [ABSTRACT FROM AUTHOR]

In this paper, a new line matching approach for high-resolution line array remote sensing images is presented. This approach establishes the correspondence of straight lines on two images by combining multiple constraints. Firstly, three geometric constraints, epipolar, direction and the point-line geometric relationship, are used in turn to reduce the number of matching candidates. After this, two similarity constraints, the double line descriptor and point-line distance, are used to determine the optimal matches. Finally, the co-linearity constraint is used to check the one-to-many and many-to-one correspondences in the results. The proposed approach is tested on eight representative image patches selected from the ZY-3 line array satellite images, and the results are compared with those of two state-of-the-art approaches. Experiments demonstrate the superiority and potential of the proposed approach due to its higher accuracy and greater number of matches in most cases. [ABSTRACT FROM AUTHOR]

Satellite hyperspectral remote sensing has gradually become an important means of Earth observation, but the existence of various types of noise seriously limits the application value of satellite hyperspectral images. With the continuous development of deep learning technology, breakthroughs have been made in improving hyperspectral image denoising algorithms based on supervised learning; however, these methods usually require a large number of clean/noisy training pairs, a target that is difficult to meet for real satellite hyperspectral imagery. In this paper, we propose a self-supervised learning-based algorithm, 3S-HSID, for denoising real satellite hyperspectral images without requiring external data support. The 3S-HSID framework can perform robust denoising of a single satellite hyperspectral image in all bands simultaneously. It first conducts a Bernoulli sampling of the input data, then uses the Bernoulli sampling results to construct the training pairs. Furthermore, the global spectral consistency and minimum local variance are used in the loss function to train the network. We use the training model to predict different Bernoulli sampling results, and the average of multiple predicted values is used as the denoising result. To prevent overfitting, we adopt a dropout strategy during training and testing. The results of denoising experiments on the simulated hyperspectral data show that the denoising performance of 3S-HSID is better than most state-of-the-art algorithms, especially in terms of maintaining the spectral characteristics of hyperspectral images. The denoising results for different types of real satellite hyperspectral data also demonstrate the reliability of the proposed method. The 3S-HSID framework provides a new technical means for real satellite hyperspectral image preprocessing. [ABSTRACT FROM AUTHOR]

The COVID-19 pandemic has affected many aspects of human life around the world, due to its tremendous outcomes on public health and socio-economic activities. Policy makers have tried to develop efficient responses based on technologies and advanced pandemic control methodologies, to limit the wide spreading of the virus in urban areas. However, techniques such as social isolation and lockdown are short-term solutions that minimize the spread of the pandemic in cities and do not invert long-term issues that derive from climate change, air pollution and urban planning challenges that enhance the spreading ability. Thus, it seems crucial to understand what kind of factors assist or prevent the wide spreading of the virus. Although AI frameworks have a very efficient predictive ability as data-driven procedures, they often struggle to identify strong correlations among multidimensional data and provide robust explanations. In this paper, we propose the fusion of a heterogeneous, spatio-temporal dataset that combine data from eight European cities spanning from 1 January 2020 to 31 December 2021 and describe atmospheric, socio-economic, health, mobility and environmental factors all related to potential links with COVID-19. Remote sensing data are the key solution to monitor the availability on public green spaces between cities in the study period. So, we evaluate the benefits of NIR and RED bands of satellite images to calculate the NDVI and locate the percentage in vegetation cover on each city for each week of our 2-year study. This novel dataset is evaluated by a tree-based machine learning algorithm that utilizes ensemble learning and is trained to make robust predictions on daily cases and deaths. Comparisons with other machine learning techniques justify its robustness on the regression metrics RMSE and MAE. Furthermore, the explainable frameworks SHAP and LIME are utilized to locate potential positive or negative influence of the factors on global and local level, with respect to our model's predictive ability. A variation of SHAP, namely treeSHAP, is utilized for our tree-based algorithm to make fast and accurate explanations. [ABSTRACT FROM AUTHOR]

A variety of remote sensing applications call for automatic optical classification of satellite images. Recently, satellite missions, such as Sentinel-2, allow us to capture images in real-time of the Earth's scenario. The classification of this large amount of data requires increasingly precise and fast methods, which must take into account not only the spectral features dependence of each individual image but also that of the temporal ones. Copulas are an excellent statistical tool, able to model joint distributions between even random variables. In this paper, we propose a new approach for Satellite Image Time-Series (SITS) land cover classification, which combines the matrix factorization to reduce the dimensionality of the data and the use of copulas distribution to model the dependencies. We will show how the use of particular copulas can improve the accuracy of classification compared to the latest methodologies used for the classification task, such as those using Neural Networks. Experiments were conducted at a study site located on Reunion Island, using Sentinel-2 SITS data. Results are compared to those achieved by several approaches commonly used to address SITS-based land cover mapping and show that the use of copulas, in combination with the matrix factorization, achieved the highest classification yield compared to competing approaches. [ABSTRACT FROM AUTHOR]

The increase in remote sensing satellite imagery with high spatial and temporal resolutions has enabled the development of a wide variety of applications for Earth observation and monitoring. At the same time, it requires new techniques that are able to manage the amount of data stored and transmitted to the ground. Advanced techniques for on-board data processing answer this problem, offering the possibility to select only the data of interest for a specific application or to extract specific information from data. However, the computational resources that exist on-board are limited compared to the ground segment availability. Alternatively, in applications such as change detection, only images containing changes are useful and worth being stored and sent to the ground. In this paper, we propose a change detection scheme that could be run on-board. It relies on a feature-based representation of the acquired images which is obtained by means of an auto-associative neural network (AANN). Once the AANN is trained, the dissimilarity between two images is evaluated in terms of the extracted features. This information can be subsequently turned into a change detection result. This study, which presents one of the first techniques for on-board change detection, yielded encouraging results on a set of Sentinel-2 images, even in light of comparison with a benchmark technique. [ABSTRACT FROM AUTHOR]

Riparian zones are dynamic ecosystems that form at the interface between the aquatic and terrestrial components of a landscape. They are shaped by complex interactions between the biophysical components of river systems, including hydrology, geomorphology, and vegetation. Remote sensing technology is a powerful tool useful for understanding riparian form, function, and change over time, as it allows for the continuous collection of geospatial data over large areas. This paper provides an overview of studies published from 1991 to 2021 that have used remote sensing techniques to map and understand the processes that shape riparian habitats and their ecological functions. In total, 257 articles were reviewed and organised into six main categories (physical channel properties; morphology and vegetation or field survey; canopy detection; application of vegetation and water indices; riparian vegetation; and fauna habitat assessment). The majority of studies used aerial RGB imagery for river reaches up to 100 km in length and Landsat satellite imagery for river reaches from 100 to 1000 km in length. During the recent decade, UAVs (unmanned aerial vehicles) have been widely used for low-cost monitoring and mapping of riverine and riparian environments. However, the transfer of RS data to managers and stakeholders for systematic monitoring as a source of decision making for and successful management of riparian zones remains one of the main challenges. [ABSTRACT FROM AUTHOR]

The importance of unsupervised clustering methods is well established in the statistics and machine learning literature. Many sophisticated unsupervised classification techniques have been made available to deal with a growing number of datasets. Due to its simplicity and efficiency in clustering a large dataset, the k-means clustering algorithm is still popular and widely used in the machine learning community. However, as with other clustering methods, it requires one to choose the balanced number of clusters in advance. This paper's primary emphasis is to develop a novel method for finding the optimum number of clusters, k, using a data-driven approach. Taking into account the cluster symmetry property, the k-means algorithm is applied multiple times to a range of k values within which the balanced optimum k value is expected. This is based on the uniqueness and symmetrical nature among the centroid values for the clusters produced, and we chose the final k value as the one for which symmetry is observed. We evaluated the proposed algorithm's performance on different simulated datasets with controlled parameters and also on real datasets taken from the UCI machine learning repository. We also evaluated the performance of the proposed method with the aim of remote sensing, such as in deforestation and urbanization, using satellite images of the Islamabad region in Pakistan, taken from the Sentinel-2B satellite of the United States Geological Survey. From the experimental results and real data analysis, it is concluded that the proposed algorithm has better accuracy and minimum root mean square error than the existing methods. [ABSTRACT FROM AUTHOR]

In space science and satellite imagery, better resolution of the data information obtained makes images clearer and interpretation more accurate. However, the huge data volume gained by the complex on-board satellite instruments becomes a problem that needs to be managed carefully. To reduce the data volume to be stored and transmitted on-ground, the signals received should be compressed, allowing a good original source representation in the reconstruction step. Image compression covers a key role in space science and satellite imagery and, recently, deep learning models have achieved remarkable results in computer vision. In this paper, we propose a spectral signals compressor network based on deep convolutional autoencoder (SSCNet) and we conduct experiments over multi/hyperspectral and RGB datasets reporting improvements over all baselines used as benchmarks and than the JPEG family algorithm. Experimental results demonstrate the effectiveness in the compression ratio and spectral signal reconstruction and the robustness with a data type greater than 8 bits, clearly exhibiting better results using the PSNR, SSIM, and MS-SSIM evaluation criteria. [ABSTRACT FROM AUTHOR]

Seagrass is an important structural and functional component of the global marine ecosystem and is of high value for its ecological services. This paper took Xincun Bay (including Xincun Harbor and Li'an Harbor) of Hainan Province as the study area, combined ground truth data, and adopted two methods to map seagrass in 2020 using Chinese GF2 satellite images: maximum-likelihood and object-oriented classification. Sentinel-2 images from 2016 to 2020 were used to extract information on seagrass distribution changes. The following conclusions were obtained. (1) Based on GF2 imagery, both the classical maximum likelihood classification (MLC) method and the object-based image analysis (OBIA) method can effectively extract seagrass information, and OBIA can also portray the overall condition of seagrass patches. (2) The total seagrass area in the study area in 2020 was about 395 hectares, most of which was distributed in Xincun Harbor. The southern coast of Xincun Harbor is an important area where seagrass is concentrated over about 228 hectares in a strip-like continuous distribution along the coastline. (3) The distribution of seagrasses in the study area showed a significant decaying trend from 2016 to 2020. The total area of seagrass decreased by 79.224 ha during the five years from 2016 to 2020, with a decay rate of 16.458%. This study is the first on the comprehensive monitoring of seagrass in Xincun Bay using satellite remote sensing images, and comprises the first use of GF2 data in seagrass research, aiming to provide a reference for remote sensing monitoring of seagrass in the South China Sea. [ABSTRACT FROM AUTHOR]

As an air pollution phenomenon, haze has become one of the focuses of social discussion. Research into the causes and concentration prediction of haze is significant, forming the basis of haze prevention. The inversion of Aerosol Optical Depth (AOD) based on remote sensing satellite imagery can provide a reference for the concentration of major pollutants in a haze, such as PM2.5 concentration and PM10 concentration. This paper used satellite imagery to study haze problems and chose PM2.5, one of the primary haze pollutants, as the research object. First, we used conventional methods to perform the inversion of AOD on remote sensing images, verifying the correlation between AOD and PM2.5. Subsequently, to simplify the parameter complexity of the traditional inversion method, we proposed using the convolutional neural network instead of the traditional inversion method and constructing a haze level prediction model. Compared with traditional aerosol depth inversion, we found that convolutional neural networks can provide a higher correlation between PM2.5 concentration and satellite imagery through a more simplified satellite image processing process. Thus, it offers the possibility of researching and managing haze problems based on neural networks. [ABSTRACT FROM AUTHOR]

In this paper, we propose a self-supervised stereo matching method based on superpixel random walk pre-matching (SRWP) and parallax-channel attention mechanism (PCAM). Our method is divided into two stages, training and testing. First, in the training stage, we obtain pre-matching results of stereo images based on superpixel random walk, and some matching points with high confidence are selected as labeled samples. Then, a stereo matching network is constructed to describe the matching correlation by calculating the attention scores of any two points between different images through the parallax-channel attention mechanism, superimposing the scores of each layer to calculate the disparity. The network is trained using the labeled samples and some unsupervised constraint criteria. Finally, in the testing stage, the trained network is used to obtain stereo matching relations of stereo images. The proposed method does not need manually labeled training samples and is more suitable for 3D reconstruction under mass satellite remote sensing data. Comparative experiments on multiple datasets show that our method has a stereo matching EPE of 2.44 and a 3D reconstruction RMSE of 2.36 m. Especially in the weak texture and parallax abrupt change regions, we can achieve more advanced performance than other methods. [ABSTRACT FROM AUTHOR]

Rivers play an important role in water supply, waterway transport, and riverine species habitations. The underwater depth of a river channel is a fundamental geometric element and a key input to studies for the aforementioned and other applications. Traditional in-situ field surveys for flow depth measurements would incur high costs and encounter technical and/or logistical difficulties, particularly for river channels of large extents, very shallow channels, and remote, inaccessible channels. Recent advances in satellite remote sensing of inland water bodies like rivers, streams, and creeks have allowed mapping flow depth in an inexpensive, convenient, and efficient manner. The purpose of this paper is to demonstrate the retrieval of river flow depth from high-resolution (1.2 m) WorldView-3 satellite imagery. The depth retrieval methods are based on the ratio of top-of-atmosphere reflectance between two pairing wavelength bands of multispectral imagery. The originality of the methods lies in using analytical relationships without resorting to ground data of river flow depth for calibration, which improves from previous studies of remote sensing of river flow depth. The methods are successfully applied to mapping the underwater bathymetry of a 26 km reach of the Nicolet River in Quebec, Canada. This study shows the importance of geometric and radiometric corrections to the satellite images. The obtained flow depths using the ratio of reflectance of the red band (630–690 nm) to that of the green band (510–580 nm), among the eight bands in the visible spectrum, agree best with in-situ measurements. This study is perhaps the first use of the analytical approach for mapping river bathymetry. It is feasible to implement the approach to other river channels, with a good potential to reduce the costs and increase the efficiency of mapping river bathymetries. [ABSTRACT FROM AUTHOR]

Green space areas are one of the key factors in people's livelihoods. Their number and size have a significant impact on both the environment and people's quality of life, including their health. Accordingly, government agencies often rely on information relating to green space areas when devising suitable plans and mandating necessary regulations. At present, obtaining information on green space areas using conventional ground surveys faces a number of limitations. This approach not only requires a lengthy period, but also tremendous human and financial resources. Given such restrictions, the status of a green space is not always up to date. Although software applications, especially those based on geographical information systems and remote sensing, have increasingly been applied to these tasks, the capability to use crowdsourcing data and produce real-time reports is lacking. This is partly because the quantity of data required has, to date, prohibited effective verification by human operators. To address this issue, this paper proposes a novel geospatial platform for green space area management by means of GIS and artificial intelligence. In the proposed system, all user-submitted data are automatically verified by deep learning classification and analyses of the greenness areas on satellite imagery. The experimental results showed that the classification and analyses can identify green space areas at accuracies of 93.50% and 97.50%, respectively. To elucidate the merits of the proposed approach, web-based application software was implemented to demonstrate multimodal data management, cleansing, and reporting. This geospatial system was thus proven to be a viable tool for assisting governmental agencies to devise appropriate plans toward sustainable development goals. [ABSTRACT FROM AUTHOR]

The EAMENA (Endangered Archaeology of the Middle East and North Africa) project is a collaboration between the Universities of Leicester, Oxford and Durham; it is funded by the Arcadia Fund and the Cultural Protection Fund. This paper explores the development of the EAMENA methodology, and discusses some of the problems of working across such a broad region. We discuss two main case studies: the World Heritage site of Cyrene illustrates how the project can use satellite imagery (dating from the 1960s to 2017), in conjunction with published data to create a detailed set of database records for a single site and, in particular, highlights the impact of modern urban expansion across the region. Conversely, the Homs Cairns case study demonstrates how the EAMENA methodology also works at an extensive scale, and integrates image interpretation (using imagery dating from the 1960s to 2016), landuse mapping and field survey (2007-2010) to record and analyse the condition of hundreds of features across a small study region. This study emphasises the impact of modern agricultural and land clearing activities. Ultimately, this paper assesses the effectiveness of the EAMENA approach, evaluating its potential success against projects using crowd-sourcing and automation for recording archaeological sites, and seeks to determine the most appropriate methods to use to document sites and assess disturbances and threats across such a vast and diverse area. [ABSTRACT FROM AUTHOR]

Anthropogenic stress and disturbance of forest ecosystems (FES) has been increasing at all scales from local to global. In rapidly changing environments, in-situ terrestrial FES monitoring approaches have made tremendous progress but they are intensive and often integrate subjective indicators for forest health (FH). Remote sensing (RS) bridges the gaps of these limitations, by monitoring indicators of FH on different spatio-temporal scales, and in a cost-effective, rapid, repetitive and objective manner. In this paper, we provide an overview of the definitions of FH, discussing the drivers, processes, stress and adaptation mechanisms of forest plants, and how we can observe FH with RS. We introduce the concept of spectral traits (ST) and spectral trait variations (STV) in the context of FH monitoring and discuss the prospects, limitations and constraints. Stress, disturbances and resource limitations can cause changes in FES taxonomic, structural and functional diversity; we provide examples how the ST/STV approach can be used for monitoring these FES characteristics. We show that RS based assessments of FH indicators using the ST/STV approach is a competent, affordable, repetitive and objective technique for monitoring. Even though the possibilities for observing the taxonomic diversity of animal species is limited with RS, the taxonomy of forest tree species can be recorded with RS, even though its accuracy is subject to certain constraints. RS has proved successful for monitoring the impacts from stress on structural and functional diversity. In particular, it has proven to be very suitable for recording the short-term dynamics of stress on FH, which cannot be cost-effectively recorded using in-situ methods. This paper gives an overview of the ST/STV approach, whereas the second paper of this series concentrates on discussing in-situ terrestrial monitoring, in-situ RS approaches and RS sensors and techniques for measuring ST/STV for FH. [ABSTRACT FROM AUTHOR]

Change detection in multi-temporal remote sensing images has usually been treated as a problem of explicitly detecting land cover transitions. To date, multi-dimensional change vector analysis has been an effective solution to such problems. However, using change vector analysis makes it hard to calculate multiple directions or kinds of change. Through combining multi-feature object-based image analysis and change vector analysis, this paper presents a novel method for object-based change detection of multiple changes. Our technique, named self-adaptive weight-change vector analysis, carries out: (1) change vector analysis to determine magnitude and direction of changes; and (2) self-adaptive weight-based analysis of the standard deviation of image objects. Furthermore, a polar representation has been adopted to acquire visual change information for image objects. This paper proposes an automatic technique that can be applied to the field of multi-feature object-based change detection for very high resolution remotely sensed images. The two-step automatic detection strategy includes extraction of changed objects using an expectation-maximization algorithm to estimate the threshold under a Gaussian assumption, and identification of different kinds of changes using a K-means clustering algorithm. The effectiveness of our approach has been tested on both multispectral and panchromatic fusion images. Results of these two experimental cases confirm that this approach can detect multiple kinds of change. We found that self-adaptive weight-change vector analysis had superior capabilities of object-based change detection compared with standard change vector analysis, yielding Kappa statistics of 0.7976 and 0.7508 for Cases 1 and 2, respectively. [ABSTRACT FROM AUTHOR]

This paper investigates the positioning accuracy of image pairs achieved by integrating images from multiple satellites. High-resolution satellite images from IKONOS, QuickBird, and KOMPSAT-2 for Daejeon, Korea were combined to produce pairs of stereo images. From single-satellite stereo pairs to multiple-satellite image pairs, all available combinations were analyzed via a rational function model (RFM). The positioning accuracy of multiple-satellite pairs was compared to a typical single-satellite stereo pair. The results show that dual-satellite integration can be an effective alternative to single-satellite stereo imagery for horizontal position mapping, but is less accurate for vertical mapping. The integration of additional higher-resolution images can improve the overall accuracy of the existing two images, but, conversely, may result in lower accuracy when very weak convergence or bisector elevation (BIE) angles occur. This highlights that the use of higher resolution images may not ensure improved accuracy, as it can result in very weak geometry. The findings confirm that multiple-satellite images can replace or enhance typical stereo pairs, but also suggest the need for careful verification, including consideration of various geometric elements and image resolution. This paper reveals the potential, limitations, and important considerations for mapping applications using images from multiple satellites. [ABSTRACT FROM AUTHOR]

Landslides represent major natural hazards, which cause every year significant loss of lives and damages to buildings, properties and lifelines. In the last decades, a significant increase in landslide frequency took place, in concomitance to climate change and the expansion of urbanized areas. Remote sensing techniques represent a powerful tool for landslide investigation: applications are traditionally divided into three main classes, although this subdivision has some limitations and borders are sometimes fuzzy. The first class comprehends techniques for landslide recognition, i.e., the mapping of past or active slope failures. The second regards landslide monitoring, which entails both ground deformation measurement and the analysis of any other changes along time (e.g., land use, vegetation cover). The third class groups methods for landslide hazard analysis and forecasting. The aim of this paper is to give an overview on the applications of remote-sensing techniques for the three categories of landslide investigations, focusing on the achievements of the last decade, being that previous studies have already been exhaustively reviewed in the existing literature. At the end of the paper, a new classification of remote-sensing techniques that may be pertinently adopted for investigating specific typologies of soil and rock slope failures is proposed. [ABSTRACT FROM AUTHOR]

Remote sensing (RS) scene classification is important for RS imagery semantic interpretation. Although tremendous strides have been made in RS scene classification, one of the remaining open challenges is recognizing RS scenes in low quality variance (e.g., various scales and noises). This paper proposes a deep salient feature based anti-noise transfer network (DSFATN) method that effectively enhances and explores the high-level features for RS scene classification in different scales and noise conditions. In DSFATN, a novel discriminative deep salient feature (DSF) is introduced by saliency-guided DSF extraction, which conducts a patch-based visual saliency (PBVS) algorithm using "visual attention" mechanisms to guide pre-trained CNNs for producing the discriminative high-level features. Then, an anti-noise network is proposed to learn and enhance the robust and anti-noise structure information of RS scene by directly propagating the label information to fully-connected layers. A joint loss is used to minimize the anti-noise network by integrating anti-noise constraint and a softmax classification loss. The proposed network architecture can be easily trained with a limited amount of training data. The experiments conducted on three different scale RS scene datasets show that the DSFATN method has achieved excellent performance and great robustness in different scales and noise conditions. It obtains classification accuracy of 98.25%, 98.46%, and 98.80%, respectively, on the UC Merced Land Use Dataset (UCM), the Google image dataset of SIRI-WHU, and the SAT-6 dataset, advancing the state-of-the-art substantially. [ABSTRACT FROM AUTHOR]

Obtaining accurate and timely land cover information is an important topic in many remote sensing applications. Using satellite image time series data should achieve high-accuracy land cover classification. However, most satellite image time-series classification methods do not fully exploit the available data for mining the effective features to identify different land cover types. Therefore, a classification method that can take full advantage of the rich information provided by time-series data to improve the accuracy of land cover classification is needed. In this paper, a novel method for time-series land cover classification using spectral, temporal, and spatial information at an annual scale was introduced. Based on all the available data from time-series remote sensing images, a refined nonlinear dimensionality reduction method was used to extract the spectral and temporal features, and a modified graph segmentation method was used to extract the spatial features. The proposed classification method was applied in three study areas with land cover complexity, including Illinois, South Dakota, and Texas. All the Landsat time series data in 2014 were used, and different study areas have different amounts of invalid data. A series of comparative experiments were conducted on the annual time-series images using training data generated from Cropland Data Layer. The results demonstrated higher overall and per-class classification accuracies and kappa index values using the proposed spectral-temporal-spatial method compared to spectral-temporal classification methods. We also discuss the implications of this study and possibilities for future applications and developments of the method. [ABSTRACT FROM AUTHOR]

Triggered by earthquakes, rainfall, or anthropogenic activities, landslides represent widespread and problematic geohazards worldwide. In recent years, multiple remote sensing techniques, including synthetic aperture radar, optical, and light detection and ranging measurements from spaceborne, airborne, and ground-based platforms, have been widely applied for the analysis of landslide processes. Current techniques include landslide detection, inventory mapping, surface deformation monitoring, trigger factor analysis and mechanism inversion. In addition, landslide susceptibility modelling, hazard assessment, and risk evaluation can be further analyzed using a synergic fusion of multiple remote sensing data and other factors affecting landslides. We summarize the 19 articles collected in this special issue of Remote Sensing of Landslide, in the terms of data, methods and applications used in the papers. [ABSTRACT FROM AUTHOR]

Accurate data about the spatial distribution and planting area of maize is important for policy making, economic development, environmental protection and food security under climate change. This paper proposes a new identification method for spring maize based on spectral and phenological features derived from the moderate resolution imaging spectroradiometer (MODIS) land surface reflectance time-series data. The method focused on the spectral differences of different land cover types in the specific phenological phases of spring maize by testing the selections and combinations of classification metrics, feature extraction methods and classifiers. Taking Liaoning province, a representative planting region of spring maize in Northeast China, as the study area, the results indicated that the combined multiple metrics, including the red reflectance, near-infrared reflectance and normalized difference vegetation index (NDVI), were conducive to the maize identification and were better than any single metric. With regard to the feature extraction and selection, maize identification based on different phenological features selected with prior knowledge was more efficient than that based on statistical features derived from the principal component analysis. Compared with the maximum likelihood classification method, the decision tree classification based on expert knowledge was more suitable for phenological features selected from some prior knowledge. In summary, discriminant rules were defined with those phenological features from multiple metrics, and the decision tree classification was used to identify maize in the study area. The producer's accuracy of maize identification was 98.57%, and the user's accuracy was 81.18%. This method can be potentially applied to an operational identification of maize at large scales based on remote sensing time-series data. [ABSTRACT FROM AUTHOR]

Earth observation satellites with various spatial, spectral and temporal resolutions provide an invaluable means for mapping and monitoring the Earth's environments. With the increasing demand of satellite images for remote and harsh environments and nature disaster areas such as earthquake, flooding, bushfires and other emergent events, quickly geo-positioning those images without using ground control points (GCPs) is much preferable and desirable. Built on the previously developed Spatial Triangulated Network (STN) concept by the first author, this paper presents a Rational Function Model (RFM) based geo-positioning method utilizing some pre-orientated image(s) as reference, instead of using GCPs. The experimental results indicate that the RFM method is more sensitive to the base-height ratio in the vertical accuracy than the physical model based geo-positioning method which was also developed by the first author. Compared to the traditional RFM based block adjustment using GCPs, the proposed RFM based method without GCP (using orientated images instead) can achieve similar accuracies when more than one orientated image, which have reasonable strong geometric relationships with the new images, are introduced into the proposed RFM based method. The proposed method is applicable to the scenarios in which geo-positioning is required for those new satellite images that only have RFM and no GCPs available, but where there exists some orientated images covering the same region. [ABSTRACT FROM AUTHOR]

Recent advances in remote sensed imagery and geospatial image processing using unmanned aerial vehicles (UAVs) have enabled the rapid and ongoing development of monitoring tools for crop management and the detection/surveillance of insect pests. This paper describes a (UAV) remote sensing-based methodology to increase the efficiency of existing surveillance practices (human inspectors and insect traps) for detecting pest infestations (e.g., grape phylloxera in vineyards). The methodology uses a UAV integrated with advanced digital hyperspectral, multispectral, and RGB sensors. We implemented the methodology for the development of a predictive model for phylloxera detection. In this method, we explore the combination of airborne RGB, multispectral, and hyperspectral imagery with ground-based data at two separate time periods and under different levels of phylloxera infestation. We describe the technology used-the sensors, the UAV, and the flight operations-the processing workflow of the datasets from each imagery type, and the methods for combining multiple airborne with ground-based datasets. Finally, we present relevant results of correlation between the different processed datasets. The objective of this research is to develop a novel methodology for collecting, processing, analysing and integrating multispectral, hyperspectral, ground and spatial data to remote sense different variables in different applications, such as, in this case, plant pest surveillance. The development of such methodology would provide researchers, agronomists, and UAV practitioners reliable data collection protocols and methods to achieve faster processing techniques and integrate multiple sources of data in diverse remote sensing applications. [ABSTRACT FROM AUTHOR]

Convolutional neural networks (CNNs) have demonstrated their ability object detection of very high resolution remote sensing images. However, CNNs have obvious limitations for modeling geometric variations in remote sensing targets. In this paper, we introduced a CNN structure, namely deformable ConvNet, to address geometric modeling in object recognition. By adding offsets to the convolution layers, feature mapping of CNN can be applied to unfixed locations, enhancing CNNs' visual appearance understanding. In our work, a deformable region-based fully convolutional networks (R-FCN) was constructed by substituting the regular convolution layer with a deformable convolution layer. To efficiently use this deformable convolutional neural network (ConvNet), a training mechanism is developed in our work. We first set the pre-trained R-FCN natural image model as the default network parameters in deformable R-FCN. Then, this deformable ConvNet was fine-tuned on very high resolution (VHR) remote sensing images. To remedy the increase in lines like false region proposals, we developed aspect ratio constrained non maximum suppression (arcNMS). The precision of deformable ConvNet for detecting objects was then improved. An end-to-end approach was then developed by combining deformable R-FCN, a smart fine-tuning strategy and aspect ratio constrained NMS. The developed method was better than a state-of-the-art benchmark in object detection without data augmentation. [ABSTRACT FROM AUTHOR]

Remote sensing is a suitable candidate for monitoring rapid changes in Polar regions, offering high-resolution spectral, spatial and radiometric data. This paper focuses on the spectral properties of dominant plant species acquired during the first week of August 2015. Twenty-eight plots were selected, which could easily be identified in the field as well as on RapidEye satellite imagery. Spectral measurements of individual species were acquired, and heavy metal contamination stress factors were measured contemporaneously. As a result, a unique spectral library of dominant plant species, heavy metal concentrations and damage ratios were achieved with an indication that species-specific changes due to environmental conditions can best be differentiated in the 1401-2400 nm spectral region. Two key arctic tundra species, Cassiope tetragona and Dryas octopetala, exhibited significant differences in this spectral region that were linked to a changing health status. Relationships between field and satellite measurements were comparable, e.g., the Red Edge Normalized Difference Vegetation Index (RENDVI) showed a strong and significant relationship (R2 = 0.82; p = 0.036) for the species Dryas octopetala. Cadmium and Lead were below detection levels while manganese, copper and zinc acquired near Longyearbyen were at concentrations comparable to other places in Svalbard. There were high levels of nickel near Longyearbyen (0.014 mg/g), while it was low (0.004 mg/g) elsewhere. [ABSTRACT FROM AUTHOR]

High-frequency (HF) radars are routinely used for remotely sensing ocean surface currents. However, the performance of the most widely used direction-finding HF radar is degraded due to the effect of the inevitable deviations of actual antenna pattern on the direction of arrival (DOA) estimation. In this paper, we quantify the DOA estimation error resulting from the deviation of the actual antenna pattern from the ideal one. Theoretical analysis and field experiment results suggest that the ratio of the deviations for the two loops dominates the DOA estimation error. Thus, eliminating the effect of the antenna pattern deviations on DOA estimation error is transformed into eliminating the effect of this ratio. From this, a calibration method based on the time-averaged local spatial coverage rate (TLSCR) is proposed to reduce the effect of the antenna pattern deviations on current extraction, which uses the ideal antenna pattern to estimate the DOA of the echoes. To validate this proposed calibration method, we assess the radar-derived radial velocities by comparing with in situ observations. The comparison results indicate that the proposed TLSCR calibration method can effectively reduce the DOA estimation error and improve the performance of the direction-finding HF radar in current observation. [ABSTRACT FROM AUTHOR]