Search

It is necessary to extract and recognize the cloud regions presented in imagery to generate satellite imagery as analysis-ready data (ARD). In this manuscript, we proposed a new deep learning model to detect cloud areas in very-high-resolution (VHR) satellite imagery by fusing two deep learning architectures. The proposed UNet3+ model with a hybrid Swin Transformer and EfficientNet (UNet3+STE) was based on the structure of UNet3+, with the encoder sequentially combining EfficientNet based on mobile inverted bottleneck convolution (MBConv) and the Swin Transformer. By sequentially utilizing convolutional neural networks (CNNs) and transformer layers, the proposed algorithm aimed to extract the local and global information of cloud regions effectively. In addition, the decoder used MBConv to restore the spatial information of the feature map extracted by the encoder and adopted the deep supervision strategy of UNet3+ to enhance the model's performance. The proposed model was trained using the open dataset derived from KOMPSAT-3 and 3A satellite imagery and conducted a comparative evaluation with the state-of-the-art (SOTA) methods on fourteen test datasets at the product level. The experimental results confirmed that the proposed UNet3+STE model outperformed the SOTA methods and demonstrated the most stable precision, recall, and F1 score values with fewer parameters and lower complexity. [ABSTRACT FROM AUTHOR]

Deep learning has shown promising performance in submeter-level mapping tasks; however, its annotation cost remains a challenge, especially when applied on a large scale. In this paper, we introduce the first submeter-level land cover mapping of Japan, employing eight classes. We present a human-in-the-loop framework that achieves national-scale mapping with a small amount of additional labeled data together with OpenEarthMap, a recently introduced benchmark dataset for global submeter-level land cover mapping. Using aerial imagery provided by the Geospatial Information Authority of Japan, we create land cover classification maps for the entire country of Japan and evaluate their accuracy. By adding a small amount of labeled data to areas where a U-Net model trained on OpenEarthMap clearly failed and retraining the model, an overall accuracy of 80% was achieved, which is a nearly 16 percentage point improvement after retraining. Our framework, with its low-cost and high-accuracy mapping results, demonstrates the potential to contribute to the automatic updating of land cover maps using submeter-level optical remote sensing data. The mapping results will be made publicly available.

Change detection (CD), as a special remote-sensing (RS) segmentation task, faces challenges, including alignment errors and illumination variation, dense small targets, and large background intraclass variance in very high-resolution (VHR) remote-sensing images. Recent methods have avoided the misjudgment caused by illumination variation and alignment errors by increasing the ability of global modeling, but the latter two problems have still not been fully addressed. In this paper, we propose a new CD model called SFCD, which increases the feature extraction capability for small targets by introducing a shifted-window (Swin) transformer. We designed a foreground-aware fusion module to use attention gates to trim low-level feature responses, enabling increased attention to the changed region compared to the background when recovering the changed region, thus reducing background interference. We evaluated our model on two CD datasets, LEVIR-CD and CDD, and obtained F1 scores of 91.78 and 97.87, respectively. The experimental results and visual interpretation show that our model outperforms several previous CD models. In addition, we adjusted the parameters and structure of the standard model to develop a lightweight version that achieves an accuracy beyond most models with only 1.55 M parameters, further validating the effectiveness of our design. [ABSTRACT FROM AUTHOR]

Satellite-derived bathymetry is increasingly attracting stakeholders’ attention tasked with remote and/or shallow depths, given its affordability compared to airborne lidar and waterborne sonar surveys. The 6-band 1.2 m Pléiades Neo (PNEO) multispectral imagery has not yet been evaluated for such a purpose. The contribution of the novel PNEO bands to the depth retrieval was assessed over unclear coastal seawaters (0.2 m−1 of vertical light attenuation in the bay of Saint-Malo, France). The relevance of the radiometric level was also tested: top-of-atmosphere (TOA) digital number (DN), TOA radiance, TOA reflectance, bottom-of-atmosphere (BOA) maritime-modeled reflectance, and BOA tropospheric-modeled reflectance. The lidar response, ranging from 0 to 20 m depth, was stratified by 90 random samples per bathymetric slice of 1 m. The model was based on an easy-to-transfer neural network (one hidden layer and three neurons). The best predictions, reaching R2test of 0.81, were equally obtained for the full PNEO dataset at TOA DN, radiance, and reflectance. For both BOA full-dataset products, the results were slightly less satisfactory: R2test of 0.75 (maritime) and 0.76 (tropospheric).

An innovative approach, mainly featured by progressive iteration of partitioning and focusing, to processing very-high-resolution spaceborne synthetic aperture radar (SAR) data is presented in this article. Due to the long integration time endured during data acquisition as well as large scene extensions being common to the advanced SAR systems, spatial dependence of range histories may be severe and must be properly dealt with. In this article, after deriving the exact two dimensional spectrum for an curved satellite orbit, we focus the entire echo data through several iterations of partitioning and focusing. By partitioning, each partitioned data block can be better focused by the following focusing step, and meanwhile, focusing with better quality enables the following partitioning with both higher granularity and less overlapped data region. Besides, a new technique aiming at eliminating spatial dependence inside a data block is also presented and deployed in the processing procedure. Owing to the feature of high granularity partitioning, the spatial dependence of many factors can be accommodated finer and easier compared with other conventional algorithms. What is more, the artifacts caused by subaperture recombination are fundamentally avoided by the fact that the partitioned blocks are not recombined until being processed into fully focused image blocks. Finally, the precision and the efficiency of the proposed methodology are validated by results on a simulated and a real SAR data.

Semantic segmentation of remote sensing images plays a crucial role in urban planning and development. How to perform automatic, fast, and effective semantic segmentation of considerable size and high-resolution remote sensing images has become the key to research. However, the existing segmentation methods based on deep learning are complex and often difficult to apply practically due to the high computational cost of the excessive parameters. In this paper, we propose an end-to-end lightweight progressive attention semantic segmentation network (LPASS-Net), which aims to solve the problem of reducing computational costs without losing accuracy. Firstly, its backbone features are based on a lightweight network, MobileNetv3, and a feature fusion network composed of a reverse progressive attentional feature fusion network work. Additionally, a lightweight non-local convolutional attention network (LNCA-Net) is proposed to effectively integrate global information of attention mechanisms in the spatial dimension. Secondly, an edge padding cut prediction (EPCP) method is proposed to solve the problem of splicing traces in the prediction results. Finally, evaluated on the public datasets BDCI 2017 and ISPRS Potsdam, the mIoU reaches 83.17% and 88.86%, respectively, with an inference time of 0.0271 s. [ABSTRACT FROM AUTHOR]

Monitoring the world's areas that are more vulnerable to natural hazards has become crucial worldwide. In order to reduce disaster risk, effective tools and relevant land cover (LC) data are needed. This work aimed to generate a high-resolution LC map of flood-prone rural villages in southwest Niger using multispectral drone imagery. The LC was focused on highly thematically detailed classes. Two photogrammetric flights of fixed-wing unmanned aerial systems (UAS) using RGB and NIR optical sensors were realized. The LC input dataset was generated using structure from motion (SfM) standard workflow, resulting in two orthomosaics and a digital surface model (DSM). The LC system is composed of nine classes, which are relevant for estimating flood-induced potential damages, such as houses and production areas. The LC was generated through object-oriented supervised classification using a random forest (RF) classifier. Textural and elevation features were computed to overcome the mapping difficulties due to the high spectral homogeneity of cover types. The training-test dataset was manually defined. The segmentation resulted in an F1_score of 0.70 and a median Jaccard index of 0.88. The RF model performed with an overall accuracy of 0.94, with the grasslands and the rocky clustered areas classes the least performant. [ABSTRACT FROM AUTHOR]

Tree that grows within a town, city and suburban areas, collection of these trees makes the urban forest. These urban forest and urban trees have impact on urban water, pollution and heat. Nowadays we are experiencing drastic climatic changes because of cutting of trees for our growth and increasing population which leads to expansion of roads, towers, and airports. Individual tree crown detection is necessary to map the forest along with feasible planning for urban areas. In this study, using WorldView-2imagery, trees in specific area are detected with object-based image analysis (OBAI) approach. Therefore with improvement in spatial and spectral resolution of an image, extracted from WorldView-2 carried out urban features with better accuracy. The aim of this research is to illustrate how object-based method can be applied to the available data to accurately find out vegetation, which can be further sub-classified to obtain area under tree canopy. The result thus obtained gives area under tree canopy with an accuracy of 92.43 % and a Kappa coefficient of 0.80. [ABSTRACT FROM AUTHOR]

Change detection (CD), as a special remote-sensing (RS) segmentation task, faces challenges, including alignment errors and illumination variation, dense small targets, and large background intraclass variance in very high-resolution (VHR) remote-sensing images. Recent methods have avoided the misjudgment caused by illumination variation and alignment errors by increasing the ability of global modeling, but the latter two problems have still not been fully addressed. In this paper, we propose a new CD model called SFCD, which increases the feature extraction capability for small targets by introducing a shifted-window (Swin) transformer. We designed a foreground-aware fusion module to use attention gates to trim low-level feature responses, enabling increased attention to the changed region compared to the background when recovering the changed region, thus reducing background interference. We evaluated our model on two CD datasets, LEVIR-CD and CDD, and obtained F1 scores of 91.78 and 97.87, respectively. The experimental results and visual interpretation show that our model outperforms several previous CD models. In addition, we adjusted the parameters and structure of the standard model to develop a lightweight version that achieves an accuracy beyond most models with only 1.55 M parameters, further validating the effectiveness of our design.

Automatic building semantic segmentation is the most critical and relevant task in several geospatial applications. Methods based on convolutional neural networks (CNNs) are mainly used in current building segmentation. The requirement of huge pixel-level labels is a significant obstacle to achieve the semantic segmentation of building by CNNs. In this paper, we propose a novel weakly supervised framework for building segmentation, which generates high-quality pixel-level annotations and optimizes the segmentation network. A superpixel segmentation algorithm can predict a boundary map for training images. Then, Superpixels-CRF built on the superpixel regions is guided by spot seeds to propagate information from spot seeds to unlabeled regions, resulting in high-quality pixel-level annotations. Using these high-quality pixel-level annotations, we can train a more robust segmentation network and predict segmentation maps. To iteratively optimize the segmentation network, the predicted segmentation maps are refined, and the segmentation network are retrained. Comparative experiments demonstrate that the proposed segmentation framework achieves a marked improvement in the building's segmentation quality while reducing human labeling efforts. [ABSTRACT FROM AUTHOR]

Building extraction using very high resolution (VHR) optical remote sensing imagery is an essential interpretation task that impacts human life. However, buildings in different environments exhibit various scales, complicated spatial distributions, and different imaging conditions. Additionally, with the spatial resolution of images increasing, there are diverse interior details and redundant context information present in building and background areas. Thus, the above-mentioned situations would create large intra-class variances and poor inter-class discrimination, leading to uncertain feature descriptions for building extraction, which would result in over- or under-extraction phenomena. In this article, a novel hierarchical disentangling network with an encoder–decoder architecture called HDNet is proposed to consider both the stable and uncertain feature description in a convolution neural network (CNN). Next, a hierarchical disentangling strategy is set up to individually generate strong and weak semantic zones using a newly designed feature disentangling module (FDM). Here, the strong and weak semantic zones set up the stable and uncertain description individually to determine a more stable semantic main body and uncertain semantic boundary of buildings. Next, a dual-stream semantic feature description is built to gradually integrate strong and weak semantic zones by the designed component feature fusion module (CFFM), which is able to generate a powerful semantic description for more complete and refined building extraction. Finally, extensive experiments are carried out on three published datasets (i.e., WHU satellite, WHU aerial, and INRIA), and the comparison results show that the proposed HDNet outperforms other state-of-the-art (SOTA) methods. [ABSTRACT FROM AUTHOR]

Urban landscapes are rapid changing ecosystems with diverse urban forms that impede the movement of organisms. Therefore, designing and modelling ecological networks to identify biodiversity reservoirs and their corridors are crucial aspects of land management in terms of population persistence and survival. However, the land cover/use maps used for landscape connectivity modelling can lack information in such a highly complex environment. In this context, remote sensing approaches are gaining interest for the development of accurate land cover/use maps. We tested the efficiency of an object-based classification using open-source projects and free images to identify vegetation strata at a very fine scale and evaluated its contribution to landscape connectivity modelling. We compared different spatial and thematic resolutions from existing databases and object-based image analyses in three French cities. Our results suggested that this remote sensing approach produced reliable land cover maps to differentiate artificial areas, tree vegetation and herbaceous vegetation. Land cover maps enhanced with the remote sensing products substantially changed the structural connectivity indices, showing an improvement up to four times the proportion of herbaceous and tree vegetation. In addition, functional connectivity indices evaluated for several forest species were mainly impacted for medium dispersers in quantitative (metrics) and qualitative (corridors) estimations. Thus, the combination of this reproductible remote sensing approach and landscape connectivity modelling at a very fine scale provides new insights into the characterisation of ecological networks for conservation planning.

For the object-based classification of high resolution remote sensing images, many people expect that introducing deep learning methods can improve then classification accuracy. Unfortunately, the input shape for deep neural networks (DNNs) is usually rectangular, whereas the shapes of the segments output by segmentation methods are usually according to the corresponding ground objects; this inconsistency can lead to confusion among different types of heterogeneous content when a DNN processes a segment. Currently, most object-based methods utilizing convolutional neural networks (CNNs) adopt additional models to overcome the detrimental influence of such heterogeneous content; however, these heterogeneity suppression mechanisms introduce additional complexity into the whole classification process, and these methods are usually unstable and difficult to use in real applications. To address the above problems, this paper proposes a simplified object-based deep neural network (SO-DNN) for very high resolution remote sensing image classification. In SO-DNN, a new segment category label inference method is introduced, in which a deep semantic segmentation neural network (DSSNN) is used as the classification model instead of a traditional CNN. Since the DSSNN can obtain a category label for each pixel in the input image patch, different types of content are not mixed together; therefore, SO-DNN does not require an additional heterogeneity suppression mechanism. Moreover, SO-DNN includes a sample information optimization method that allows the DSSNN model to be trained using only pixel-based training samples. Because only a single model is used and only a pixel-based training set is needed, the whole classification process of SO-DNN is relatively simple and direct. In experiments, we use very high-resolution aerial images from Vaihingen and Potsdam from the ISPRS WG II/4 dataset as test data and compare SO-DNN with 6 traditional methods: O-MLP, O+CNN, OHSF-CNN, 2-CNN, JDL and U-Net. Compared with the best-performing method among these traditional methods, the classification accuracy of SO-DNN is improved by up to 7.71% and 10.78% for single images from Vaihingen and Potsdam, respectively, and the average classification accuracy is improved by 2.46% and 2.91% for the Vaihingen and Potsdam images, respectively. SO-DNN relies on fewer models and easier-to-obtain samples than traditional methods, and its stable performance makes SO-DNN more valuable for practical applications. [ABSTRACT FROM AUTHOR]

Semantic segmentation of remote sensing images plays a crucial role in urban planning and development. How to perform automatic, fast, and effective semantic segmentation of considerable size and high-resolution remote sensing images has become the key to research. However, the existing segmentation methods based on deep learning are complex and often difficult to apply practically due to the high computational cost of the excessive parameters. In this paper, we propose an end-to-end lightweight progressive attention semantic segmentation network (LPASS-Net), which aims to solve the problem of reducing computational costs without losing accuracy. Firstly, its backbone features are based on a lightweight network, MobileNetv3, and a feature fusion network composed of a reverse progressive attentional feature fusion network work. Additionally, a lightweight non-local convolutional attention network (LNCA-Net) is proposed to effectively integrate global information of attention mechanisms in the spatial dimension. Secondly, an edge padding cut prediction (EPCP) method is proposed to solve the problem of splicing traces in the prediction results. Finally, evaluated on the public datasets BDCI 2017 and ISPRS Potsdam, the mIoU reaches 83.17% and 88.86%, respectively, with an inference time of 0.0271 s.

Satellite picture characterization process includes gathering the picture pixel esteems into significant classifications. A few satellite picture characterization strategies and methods are accessible. In existing Markov random field (MRF) is utilized for grouping the satellite information, with this technique not ready to bunch precisely all the classes. In our proposed strategy self-sorting out maps as a bunching method is utilized. Self-sorting out maps figure out how to bunch information dependent on comparability, topology, with an inclination of doling out a similar number of examples to each class. Self- sorting out maps are utilized both to group information and to lessen the dimensionality of information. They are roused by the tactile and engine mappings in the vertebrate cerebrum, which additionally appear to naturally arranging data topologically. Group Classifiers merge results from numerous feeble students into one highcaliber group model. [ABSTRACT FROM AUTHOR]

Urban trees are valuable in, inter alia, ameliorating air pollution and mitigating the effects associated with urban heat islands. The dearth of tree cover maps is a major challenge for urban planners in the management of urban trees. This work adopts remote sensing approaches to provide urban tree cover maps which can strengthen urban landscape management. Whereas traditional pixel-based classification approaches have been commonly used in image classification, they are not well-suited for urban tree mapping due to their failure to fully explore the image's spatial and spectral characteristics. Object-based classification techniques produce improved accuracies using additional variables. This study depicts the capability of object-based image analysis (OBIA) in mapping common urban trees using very high-resolution (VHR) WorldView-2 (WV-2) imagery. The study tests the utility of WV-2 bands and other feature variables in the object-based mapping of common urban trees and other land cover classes. Furthermore, the study compares the utility of Support Vector Machine (SVM) and Random Forest (RF) in the object-based mapping of common urban trees and other land cover classes. The results show that the Normalized Difference Vegetation Index (NDVI), NIR 1 and NIR 2 bands were important in the classification of common urban trees and other land cover classes. The RF classifier performed better than SVM, with an overall accuracy of 91.9% as compared to 87.3% for SVM. The results of this study offer insight to urban authorities with knowledge on the segmentation parameters, classification methods and feature variables for mapping urban trees, valuable in urban tree management. [ABSTRACT FROM AUTHOR]

This paper introduces some of the image processing techniques developed in the Canada Research Chair in Advanced Geomatics Image Processing Laboratory (CRC-AGIP Lab) and in the Department of Geodesy and Geomatics Engineering (GGE) at the University of New Brunswick (UNB), Canada. The techniques were developed by innovatively/“smartly” utilizing the characteristics of the available very high resolution optical remote sensing images to solve important problems or create new applications in photogrammetry and remote sensing. The techniques to be introduced are: automated image fusion (UNB-PanSharp), satellite image online mapping, street view technology, moving vehicle detection using single set satellite imagery, supervised image segmentation, image matching in smooth areas, and change detection using images from different viewing angles. Because of their broad application potential, some of the techniques have made a global impact, and some have demonstrated the potential for a global impact.

Monitoring the world’s areas that are more vulnerable to natural hazards has become crucial worldwide. In order to reduce disaster risk, effective tools and relevant land cover (LC) data are needed. This work aimed to generate a high-resolution LC map of flood-prone rural villages in southwest Niger using multispectral drone imagery. The LC was focused on highly thematically detailed classes. Two photogrammetric flights of fixed-wing unmanned aerial systems (UAS) using RGB and NIR optical sensors were realized. The LC input dataset was generated using structure from motion (SfM) standard workflow, resulting in two orthomosaics and a digital surface model (DSM). The LC system is composed of nine classes, which are relevant for estimating flood-induced potential damages, such as houses and production areas. The LC was generated through object-oriented supervised classification using a random forest (RF) classifier. Textural and elevation features were computed to overcome the mapping difficulties due to the high spectral homogeneity of cover types. The training-test dataset was manually defined. The segmentation resulted in an F1_score of 0.70 and a median Jaccard index of 0.88. The RF model performed with an overall accuracy of 0.94, with the grasslands and the rocky clustered areas classes the least performant.

Very high-resolution satellite imagery from the latest generation commercial platforms provides an unprecedented capacity for imaging the Earth with very high spatial detail. However, these data are generally expensive, particularly if large areas or temporal sequences are required. In recent years, lower quality imagery has been enabled through the launch of constellations of small satellites with short revisit time. In this article, we apply for the first time a statistical approach to downscale and bias-correct these multispectral satellite data using the information contained in a limited training set of very high-resolution images. The technique, based on the direct sampling algorithm, aims at extending the coverage of high-resolution images by sampling data from a training data set, where similar lower resolution data patterns are found. Unlike the majority of the current downscaling techniques, the approach does not require colocated fine-resolution data, but it is based on the use of training images similar to the target zone. A novel specific setup is proposed, which is adaptive to different types of landscapes with no additional user effort. The results show that the proposed technique can generate more realistic images than the traditional approaches based on the parametric bias correction and bicubic interpolation. In particular, properties such as the intensity histogram, spatial correlation, and connectivity are accurately preserved. The proposed approach can be used to extend the footprint of the high-resolution images to generate new time frames or to downscale the remote sensing imagery based on a distant but structurally similar training image. [ABSTRACT FROM AUTHOR]

Automatic building semantic segmentation is the most critical and relevant task in several geospatial applications. Methods based on convolutional neural networks (CNNs) are mainly used in current building segmentation. The requirement of huge pixel-level labels is a significant obstacle to achieve the semantic segmentation of building by CNNs. In this paper, we propose a novel weakly supervised framework for building segmentation, which generates high-quality pixel-level annotations and optimizes the segmentation network. A superpixel segmentation algorithm can predict a boundary map for training images. Then, Superpixels-CRF built on the superpixel regions is guided by spot seeds to propagate information from spot seeds to unlabeled regions, resulting in high-quality pixel-level annotations. Using these high-quality pixel-level annotations, we can train a more robust segmentation network and predict segmentation maps. To iteratively optimize the segmentation network, the predicted segmentation maps are refined, and the segmentation network are retrained. Comparative experiments demonstrate that the proposed segmentation framework achieves a marked improvement in the building’s segmentation quality while reducing human labeling efforts.

Building extraction using very high resolution (VHR) optical remote sensing imagery is an essential interpretation task that impacts human life. However, buildings in different environments exhibit various scales, complicated spatial distributions, and different imaging conditions. Additionally, with the spatial resolution of images increasing, there are diverse interior details and redundant context information present in building and background areas. Thus, the above-mentioned situations would create large intra-class variances and poor inter-class discrimination, leading to uncertain feature descriptions for building extraction, which would result in over- or under-extraction phenomena. In this article, a novel hierarchical disentangling network with an encoder–decoder architecture called HDNet is proposed to consider both the stable and uncertain feature description in a convolution neural network (CNN). Next, a hierarchical disentangling strategy is set up to individually generate strong and weak semantic zones using a newly designed feature disentangling module (FDM). Here, the strong and weak semantic zones set up the stable and uncertain description individually to determine a more stable semantic main body and uncertain semantic boundary of buildings. Next, a dual-stream semantic feature description is built to gradually integrate strong and weak semantic zones by the designed component feature fusion module (CFFM), which is able to generate a powerful semantic description for more complete and refined building extraction. Finally, extensive experiments are carried out on three published datasets (i.e., WHU satellite, WHU aerial, and INRIA), and the comparison results show that the proposed HDNet outperforms other state-of-the-art (SOTA) methods.

Semi-arid parkland agrosystems are strongly sensitive to climate change and anthropic pressure. In the context of sustainability research, trees are considered critical for various ecosystem services covering environment quality as well as food security and health. But their actual ecological impact on both cropland and natural vegetation is not well understood yet, and collecting spatial and structural information around agroforestry systems is becoming an important issue. Tree mapping in semi-arid parklands could be one of these prerequisites. While for obtaining an exhaustive inventory of individual trees and for analysing their spatial distribution, remote sensing is the ideal tool. However, it has been noted that depending on the spatial resolution and sensor spectral characteristics, tree species cannot be distinguished clearly, even in the sparsely vegetated semi-arid ecosystems of West Africa. Thus, this work focuses on assessing the capabilities of Worldview-3 imagery, acquired in 8 spectral bands, to detect, delineate, and identify certain key tree species in the Faidherbia albida parkland in Bambey, Senegal, based on a ground-truth database corresponding to 5000 trees. The tree crowns are delineated through NDVI thresholding and consecutive filtering to provide object-based radiometric signatures, radiometric indices, and textural information. A factorial discriminant analysis is then performed, which indicates that only four out of the seven most abundant species in the study area can be discriminated: “Faidherbia albida”,” Azadirachta indica”, “Balanites aegyptiaca” and “Tamarindus indica”. Next, random forest and support vector machine classifiers are employed to identify the optimal combination of classifier parameters to discriminate these classes with a high accuracy, robustness, and stability. The linear support vector machine with cost=1 and gamma=0.01 provides the optimal results with a global accuracy of 88 % and kappa of 0.71. This classifier is applied to the whole study area to map all the trees with crowns larger than 2 m, sorted in four identified species and a fifth common group of unidentified species. This map thus enables analysing the variability in tree density and the spatial distribution of different species. Such information can afterwards be correlated to the ecological functioning of the parkland and local practices, and offers promising opportunities to help future sustainability initiatives in different socio-ecological contexts.

With the popularization of very high resolution polarimetric synthetic aperture radar image dataset, it is essential to re-investigate the classification scheme for 2-D land cases. The Touzi scattering vector model, a unique and roll-invariant decomposition solution, is employed to extract the scattering properties of different land covers. The parameters of Touzi decomposition act as input dataset for initial classification. A novel classifying algorithm is put forward by means of integrating the Touzi decomposition with conventional Wishart statistical models. Quantitative experiments are then conducted using uninhabited aerial vehicle synthetic aperture radar sample data for evaluating the performance of this new proposed approach. It can be concluded from the experimental results that the new proposed method is superior to the classical method in terms of producer accuracy, user accuracy, and overall accuracy. [ABSTRACT FROM AUTHOR]

Gas hydrate (GH) systems constitute methane sinks sensitive to environmental changes such as pressure, temperature, and salinity. It remains a matter of debate as to whether the large GH system of the Black Sea has reached a steady state since the last glacial maximum (LGM). We report on an irregular free gas distribution in specific sediment layers marking an irregular bottom‐simulating reflector (BSR). This anomalous free gas distribution revealed by very high resolution seismic images, acquired by a deep‐towed multichannel seismic system, might be evidence of an ongoing migration of the base of the GH stability zone (GHSZ). We show that the reequilibrium is not occurring homogeneously as overpressure from hydrate dissociation slows their decomposition in specific sedimentary layers. The Black Sea example highlights that dissociation and the associated methane release in the water column or even in the atmosphere could be largely delayed by overpressure accumulation. Plain Language Summary: Methane hydrate is an ice‐like compound composed of a cage of water molecules enclosing a methane molecule. Hydrates can form where water and methane are present under high pressure and low temperatures, for example, in deep‐sea sediments. As a result of climate change (e.g., seawater temperature increase), hydrates can melt and release free gas and water. Yet we observe that hydrates are present where they should have melted according to modeling. We explain this irregular melting by differing properties of the host sediments and different quantities of hydrate in the sediments. Methane in the Earth's atmosphere is a strong greenhouse gas. The release of methane from hydrate melting has been proposed as a runaway process where the methane released increases global warming, which further increases hydrate melting and methane release, repeating the cycle. Our results show that the destabilization of a hydrate system is actually a slow process, spanning several millennia. As such, a catastrophic destabilization of a gas hydrate system is unlikely. Key Points: Very high resolution (<1 m) deep‐towed seismic imaging of the bottom‐simulating reflector allows characterizing gas hydrates dynamicsIrregular bottom‐simulating reflector is an indicator of the transient state of a gas hydrate system in the Black SeaPore overpressure and hydrate recrystallization explain the long process of the readjustment of the base of the gas hydrate stability zone [ABSTRACT FROM AUTHOR]

The regular inspection of the crane tracks of storage cranes at the Container Terminal Altenwerder (CTA), Hamburg requires high accuracy of measurements to determine its position. The allowed tolerances are in the range of 10 mm in the XY plane on a track length of 300 m. The traditional semi-automatic surveying methods are slow and require the interruption of the activities in the storage blocks. The research project AeroInspekt proposed a fully automatic measurement of the position of the tracks using UAV-based photogrammetry. In this paper, the results of the test campaign, carried out in June 2020, were presented where different cameras (150 mm and 80 mm telelens) and flight speeds (1.1 m/s and 1.9 m/s) at a 35 m flying height were performed. Furthermore, an automated rail delineation in the derived surface model was developed and evaluated with ground reference measurements. The results show that the required accuracy of the rail position with an RMSE of 3 mm in XY plane and 8 mm in altitude can be achieved with comparatively less disruption of regular block activities.

This study aims to assess the capabilities of a weather forecasting system based on simulations performed with the COSMO (COnsortium for Small-scale Modeling) model over a domain located in southern Italy, employing a spatial resolution of about 1 km, driven by ECMWF-IFS global data. The model is run daily at the Italian Aerospace Research Center (CIRA), and the evaluation was performed from January to May 2018 using a combination of observational data, specifically data provided by the CIRA meteorological station, wind profiler and ceilometer. Moreover, data provided by radio sounding located at Pratica di Mare and ground stations at two other locations were also used. A model configuration optimized through a tuning procedure over the domain considered was employed, while the evaluation was performed by comparing daily values of several variables and using standard monitoring indices. The results highlight that the model has good capability in reproducing daily values of temperature, while precipitation intensity is generally underestimated, even if rain patterns are well captured (alternating rainy and dry days). Good agreement is also reported for wind speed, especially at 100 and 500 m altitude. Regarding radio sounding data, the COSMO model configuration selected can reproduce the vertical profile of temperature and dew point, with the exception of inversion points. Evaluation against ceilometer data is achieved in terms of cloud height and planetary boundary layer height.

During the last decades, archaeological site looting throughout Iraq has increased significantly up to a point where some of the most famous and relevant ancient Mesopotamian cities are currently threatened in their integrity. Several important archaeological monuments and artifacts have been destroyed, due to ISIL attacks and associated looting. Since 2016, the policies of the European Union have been increasingly harsh to condemn these atrocious acts of destruction. In such a scenario, the European Union Satellite Centre can be an invaluable instrument for the identification and assessment of the damage in areas occupied by ISIL. A detailed view of the damage suffered by the Nineveh and Nebi Yunus ancient sites, in Iraq, was assessed via visual inspection. The analysis was conducted considering the main events that occurred in the city of Mosul, between November 2013 and March 2018. More than 25 satellite images, new acquisitions and archived, supported by collateral data, allowed the detection and classification of the damage occurred over time. A description of the methodology and the classification of category and type of damage is presented. The results of the analysis confirm the dramatic levels of destruction that these two ancient sites have been suffering since 2013. The analysis reported in this paper is part of a wider study that the SatCen conducted in cooperation with the EU Counter-Terrorism Office and PRISM Office. The whole activity aimed at confirming to EU institutions the massive looting and trafficking operated in the area. The results have been provided to archaeologists in the field as well in support of local authorities who are trying to evaluate the current situation in the area.

Global food production in the developing world occurs within sub-hectare fields that are difficult to identify with moderate resolution satellite imagery. Knowledge about the distribution of these fields is critical in food security programs. We developed a semi-automated image segmentation approach using wall-to-wall sub-meter imagery with high-performance computing to map crop area (CA) throughout Tigray, Ethiopia that encompasses over 41,000 km 2 . Multiple processing streams were tested to minimize mapping error while applying five unique smoothing kernels to capture differences in land surface texture associated to CA. Typically, very-small fields (mean < 2 ha) have a smooth image roughness compared to natural scrub/shrub woody vegetation at the ~1 m scale and these features can be segmented in panchromatic imagery with multi-level histogram thresholding. Multi-temporal very-high resolution (VHR) panchromatic imagery with multi-spectral VHR are sufficient in extracting critical CA information needed in food security programs. A 2011 to 2015 CA map was produced, using over 3000 WorldView-1 panchromatic images wall-to-wall in 1/2° mosaics for Tigray, Ethiopia. CA was evaluated with nearly 3000 WorldView-2 2 m multispectral 250 × 250 m image subsets by seven expert interpretations, and with in-situ global positioning system photography. CA estimates ranged from 32 to 41% in sub regions of Tigray with median maximum per bin commission and omission errors of 11% and 1% respectively, with most of the error occurring in bins <15%. This empirical, simple, and low direct cost approach via U.S. government license agreement to access commercial VHR data, could be a viable big-data high-performance computing methodology to extract wall-to-wall CA for other regions of the world that have very-small agriculture fields with similar image texture. [ABSTRACT FROM AUTHOR]

Detecting urban areas from very high resolution (VHR) remote sensing images plays an important role in the field of Earth observation. The recently-developed deep convolutional neural networks (DCNNs), which can extract rich features from training data automatically, have achieved outstanding performance on many image classification databases. Motivated by this fact, we propose a new urban area detection method based on DCNNs in this paper. The proposed method mainly includes three steps: (i) a visual dictionary is obtained based on the deep features extracted by pre-trained DCNNs; (ii) urban words are learned from labeled images; (iii) the urban regions are detected in a new image based on the nearest dictionary word criterion. The qualitative and quantitative experiments on different datasets demonstrate that the proposed method can obtain a remarkable overall accuracy (OA) and kappa coefficient. Moreover, it can also strike a good balance between the true positive rate (TPR) and false positive rate (FPR). [ABSTRACT FROM AUTHOR]

Very high resolution (VHR) remote sensing imagery has been used for land cover classification, and it tends to a transition from land-use classification to pixel-level semantic segmentation. Inspired by the recent success of deep learning and the filter method in computer vision, this work provides a segmentation model, which designs an image segmentation neural network based on the deep residual networks and uses a guided filter to extract buildings in remote sensing imagery. Our method includes the following steps: first, the VHR remote sensing imagery is preprocessed and some hand-crafted features are calculated. Second, a designed deep network architecture is trained with the urban district remote sensing image to extract buildings at the pixel level. Third, a guided filter is employed to optimize the classification map produced by deep learning; at the same time, some salt-and-pepper noise is removed. Experimental results based on the Vaihingen and Potsdam datasets demonstrate that our method, which benefits from neural networks and guided filtering, achieves a higher overall accuracy when compared with other machine learning and deep learning methods. The method proposed shows outstanding performance in terms of the building extraction from diversified objects in the urban district. [ABSTRACT FROM AUTHOR]

Up-to-date catalogs of the urban tree population are of importance for municipalities to monitor and improve quality of life in cities. Despite much research on automation of tree mapping, mainly relying on dedicated airborne LiDAR or hyperspectral campaigns, tree detection and species recognition is still mostly done manually in practice. We present a fully automated tree detection and species recognition pipeline that can process thousands of trees within a few hours using publicly available aerial and street view images of Google Maps TM . These data provide rich information from different viewpoints and at different scales from global tree shapes to bark textures. Our work-flow is built around a supervised classification that automatically learns the most discriminative features from thousands of trees and corresponding, publicly available tree inventory data. In addition, we introduce a change tracker that recognizes changes of individual trees at city-scale, which is essential to keep an urban tree inventory up-to-date. The system takes street-level images of the same tree location at two different times and classifies the type of change (e.g., tree has been removed). Drawing on recent advances in computer vision and machine learning, we apply convolutional neural networks (CNN) for all classification tasks. We propose the following pipeline: download all available panoramas and overhead images of an area of interest, detect trees per image and combine multi-view detections in a probabilistic framework, adding prior knowledge; recognize fine-grained species of detected trees. In a later, separate module, track trees over time, detect significant changes and classify the type of change. We believe this is the first work to exploit publicly available image data for city-scale street tree detection, species recognition and change tracking, exhaustively over several square kilometers, respectively many thousands of trees. Experiments in the city of Pasadena, California, USA show that we can detect >70% of the street trees, assign correct species to >80% for 40 different species, and correctly detect and classify changes in >90% of the cases. [ABSTRACT FROM AUTHOR]