High Resolution Optical Imagery Research Articles

Research on Building Target Detection Based on High-Resolution Optical Remote Sensing Imagery

High-resolution remote sensing image building target detection has wide application value in the fields of land planning, geographic monitoring, smart cities and other fields. However, due to the complex background of remote sensing imagery, some detailed features of building targets are less distinguishable from the background. When carrying out the detection task, it is prone to problems such as distortion and the missing of the building outline. To address this challenge, we developed a novel building target detection method. First, a building detection method based on rectangular approximation and region growth was proposed, and a saliency detection model based on the foreground compactness and local contrast of manifold ranking is used to obtain the saliency map of the building region. Then, the boundary prior saliency detection method based on the improved manifold ranking algorithm was proposed for the target area of buildings with low contrast with the background in remote sensing imagery. Finally, fusing the results of the rectangular approximation-based and saliency-based detection, the proposed fusion method improved the Recall and F1 value of building detection, indicating that this paper provides an effective and efficient high-resolution remote sensing image building target detection method.

A Multi-Scale Spatial Attention Region Proposal Network for High-Resolution Optical Remote Sensing Imagery

Deep convolutional neural networks (DCNNs) are driving progress in object detection of high-resolution remote sensing images. Region proposal generation, as one of the key steps in object detection, has also become the focus of research. High-resolution remote sensing images usually contain various sizes of objects and complex background, small objects are easy to miss or be mis-identified in object detection. If the recall rate of region proposal of small objects and multi-scale objects can be improved, it will bring an improvement on the performance of the accuracy in object detection. Spatial attention is the ability to focus on local features in images and can improve the learning efficiency of DCNNs. This study proposes a multi-scale spatial attention region proposal network (MSA-RPN) for high-resolution optical remote sensing imagery. The MSA-RPN is an end-to-end deep learning network with a backbone network of ResNet. It deploys three novel modules to fulfill its task. First, the Scale-specific Feature Gate (SFG) focuses on features of objects by processing multi-scale features extracted from the backbone network. Second, the spatial attention-guided model (SAGM) obtains spatial information of objects from the multi-scale attention maps. Third, the Selective Strong Attention Maps Model (SSAMM) adaptively selects sliding windows according to the loss values from the system’s feedback, and sends the windowed samples to the spatial attention decoder. Finally, the candidate regions and their corresponding confidences can be obtained. We evaluate the proposed network in a public dataset LEVIR and compare with several state-of-the-art methods. The proposed MSA-RPN yields a higher recall rate of region proposal generation, especially for small targets in remote sensing images.

ASSESSING CROPLAND ABANDONMENT FROM VIOLENT CONFLICT IN CENTRAL MALI WITH SENTINEL-2 AND GOOGLE EARTH ENGINE

Abstract. The proposed analysis based on Sentinel-2 imagery provides evidence of impacts of the conflict in the Mopti region (central Mali), which has led to widescale cropland abandonment. This area is characterized by rapidly rising levels of violence since 2018, due to the presence of armed groups and the proliferation of self-defence militias. This study investigates how high-resolution optical imagery can help evaluate the linkages between violence and land cover / land use (LCLU) change. The processing environment of Google Earth Engine was used to generate the so-called 3-Period TimeScan (3PTS) product, a RGB composite combining the maximum NDVI values in the beginning, in the middle and in the end of the growing season, used to single out cultivated land for each year of interest. Theoretically, the period between June 15th and October 15th covers an annual agricultural cycle for the considered area; consequently, images acquired during that period were used to generate the 3PTS composites for the year of interest (2019) and for pre-conflict years. By comparing the situations before and after the start of the crisis, each populated site was categorized according to the degree of cropland change detected in its surroundings. The resulting overview map enables a regional-scale interpretation of farming activities in 2019, clearly highlighting localized areas of cropland abandonment in the region and showing a strong spatial correlation with incidence of conflict.

The Approach for Studying Variability of Sea Wave Spectra in a Wide Range of Wavelengths from High-Resolution Satellite Optical Imagery

The development and validation of a method for remote measurement of the spectra of sea waves, which significantly expands the capabilities to study surface waves in large water areas in a wide range of wavelengths, is described. The applied approach is based on the use of retrieving operators, which are constructed by the method of numerical simulation, taking into account the nonlinear nature of brightness field modulation by the slopes of the sea surface. Retrieving operators have a set of parameters that are adapted to the real conditions of aerospace imaging. To assess the adequacy of the retrieving of wave spectra recorded from satellite images, they are compared with the spectra obtained by ground-based means under controlled conditions. The studies have shown the adequacy of remote measurement of the spectra of slopes and elevations of sea waves with wavelengths in the range 0.1–1 m. The possibility of using the developed method for studying the variability of sea waves in the coastal zone under conditions of limited fetch, including in the presence of anthropogenic disturbances, is shown.

Remote sensing techniques: mapping and monitoring of mangrove ecosystem\u2014a review

Mangrove forests are considered to be the most productive ecosystem yet vanishing rapidly over the world. They are mostly found in the intertidal zone and sheltered by the seacoast. Mangroves have potential socio-economic benefits such as protecting the shoreline from storm and soil erosion, flood and flow control, acting as a carbon sink, provides a fertile breeding ground for marine species and fauna. It also acts as a source of income by providing various forest products. Restoration and conservation of mangrove forests remain a big challenge due to the large and inaccessible areas covered by mangroves forests which makes field assessment difficult and time-consuming. Remote sensing along with various digital image classification approaches seem to be promising in providing better and accurate results in mapping and monitoring the mangroves ecosystem. This review paper aims to provide a comprehensive summary of the work undertaken, and addresses various remote sensing techniques applied for mapping and monitoring of the mangrove ecosystem, and summarize their potential and limitation. For that various digital image classification techniques are analyzed and compared based on the type of image used with its spectral resolution, spatial resolution, and other related image features along with the accuracy of the classification to derive specific class information related to mangroves. The digital image classification techniques used for mangrove mapping and monitoring in various studies can be classified into pixel-based, object-based, and knowledge-based classifiers. The various satellite image data analyzed are ranged from light detection and ranging (LiDAR), hyperspectral and multispectral optical imagery, synthetic aperture radar (SAR), and aerial imagery. Supervised state of the art machine learning/deep machine learning algorithms which use both pixel-based and object-based approaches and can be combined with the knowledge-based approach are widely used for classification purpose, due to the recent development and evolution in these techniques. There is a huge future scope to study the performance of these classification techniques in combination with various high spatial and spectral resolution optical imageries, SAR and LiDAR, and also with multi-sensor, multiresolution, and temporal data.

Forest Aboveground Biomass Estimation and Mapping through High-Resolution Optical Satellite Imagery—A Literature Review

This paper provides a comprehensive literature review on forest aboveground biomass (AGB) estimation and mapping through high-resolution optical satellite imagery (≤5 m spatial resolution). Based on the literature review, 44 peer-reviewed journal articles were published in 15 years (2004–2019). Twenty-one studies were conducted in Asia, eight in North America and Africa, five in South America, and four in Europe. This review article gives a glance at the published methodologies for AGB prediction modeling and validation. The literature review suggested that, along with the integration of other sensors, QuickBird, WorldView-2, and IKONOS satellite images were most widely used for AGB estimations, with higher estimation accuracies. All studies were grouped into six satellite-derived independent variables, including tree crown, image textures, tree shadow fraction, canopy height, vegetation indices, and multiple variables. Using these satellite-derived independent variables, most of the studies used linear regression (41%), while 30% used linear multiple regression and 18% used non-linear (machine learning) regression, while very few (11%) studies used non-linear (multiple and exponential) regression for estimating AGB. In the context of global forest AGB estimations and monitoring, the advantages, strengths, and limitations were discussed to achieve better accuracy and transparency towards the performance-based payment mechanism of the REDD+ program. Apart from technical limitations, we realized that very few studies talked about real-time monitoring of AGB or quantifying AGB change, a dimension that needs exploration.

SUPERVISED CLASSIFICATION METHODS FOR AUTOMATIC DAMAGE DETECTION CAUSED BY HEAVY RAINFALL USING MULTITEMPORAL HIGH RESOLUTION OPTICAL IMAGERY AND AUXILIARY DATA

Abstract. In the context of climate change and rising frequency of extreme hydro-meteorological events around the world, flood risk management and mapping of heavy rainfall-related damages represent an ongoing critical challenge. For decades now, remote sensing has been largely used to investigate spatial and temporal changes in land use and water resources. Today, different satellite products provide fast and crucial knowledge for the study of hydrological disasters over large areas, possibly in remote regions, with high spatial resolution and high revisit frequency. Yet, until now, few works have sought to detect the full range of extreme rainfall-related damages with optical imagery, especially those caused by intense rainwater runoff beyond the direct vicinity of major waterways. The work presented in this paper focuses on the Aude severe weather event of October 15th, 2018, in the South of France, for which more than a thousand claims for agricultural disaster were registered, both related to river overflowing and rainwater runoff.The full resources of ground truths, contextual information, land use as well as digital elevation model (DEM) combined to high resolution and high frequency optical imagery (Sentinel-2, Pléiades) are used to develop an automatic damage detection method based on supervised classification algorithms. Through the combination of several indicators characterizing heterogeneous spectral variations among agricultural plots following the event, a Gaussian process classifier achieved various classification accuracies up to 90% on a large comparable and independent photo-interpreted validation sample. This work builds great expectations for applications in other areas with contrasted climate, topography and land cover.

The 2020 Hpakant Jade Mine Disaster, Myanmar: A multi-sensor investigation for slope failure

A quarry failure along the slopes of the Wai Khar open-pit jade mine in Hpakant, Myanmar has led to the deaths of at least 172 jade miners on 2 July 2020. This paper conducts a systematic investigation of the incident by integrating data from multiple sensors, including high-resolution optical imagery, Sentinel-1 synthetic aperture radar (SAR) images, unmanned aerial system (UAS) footage, SRTM and ALOS digital elevation models (DEMs), soil moisture product from multi-spectral Landsat-8 satellite and precipitation records from the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS). Optical imagery, UAS footage and DEMs allow us to build a comprehensive mapping of tailing areas and quarry scarps from 2010 and reconstruct the 2D pit geometry prior to failure. Deformation signals from multi-temporal SAR interferometry (MTInSAR), soil moisture variations and precipitation trends further allow us to identify possible failure causes. To evaluate the quality of deformation obtained from different distributed-scatterer phase estimators, we develop an empirical mapping function based on areal fraction values to facilitate the comparison of temporal coherence values that are differently formulated in each phase estimator. The comparison shows that phase linking algorithm outperforms the small baseline subset method in terms of signal recovery and phase reliability. Our investigation points out that the mining site is under aggressive mining cycles that are exacerbated by frequent, uncontrolled landslides. Seepage failure, which involves the expulsion of water from rapidly compacting tailings, may be a critical factor in the 2020 incident. Instead of extreme weather, the failure had occurred under normal to drier conditions. This means that the sliding planes were already in a critical state, which is evident from the accelerated deformation around the collapse area since the beginning of 2020. Based on these findings, we provide recommendations to improve mining site regulations and management practices for safer open-pit mining in Myanmar and probably in similar contexts outside Myanmar.

Gray level co-occurrence matrix (GLCM) texture based crop classification using low altitude remote sensing platforms.

Crop classification in early phenological stages has been a difficult task due to spectrum similarity of different crops. For this purpose, low altitude platforms such as drones have great potential to provide high resolution optical imagery where Machine Learning (ML) applied to classify different types of crops. In this research work, crop classification is performed at different phenological stages using optical images which are obtained from drone. For this purpose, gray level co-occurrence matrix (GLCM) based features are extracted from underlying gray scale images collected by the drone. To classify the different types of crops, different ML algorithms including Random Forest (RF), Naive Bayes (NB), Neural Network (NN) and Support Vector Machine (SVM) are applied. The results showed that the ML algorithms performed much better on GLCM features as compared to gray scale images with a margin of 13.65% in overall accuracy.

Adaptive region-based post-classification framework for land-cover mapping improvement using very high spatial resolution optical imagery

Land cover mapping using remote sensing optical images with a very high spatial resolution (VHSR) plays an important role in observing the Earth’s surface. However, classification maps are usually affected by salt-and-pepper noise because VHSR optical images usually have a low resolution of ground targets in terms of spectral reflectance. An adaptive region-based post-classification framework (ARPF) is proposed for improving the initial classification map while using a VHSR optical image to further smooth the noise of initial classified maps. First, different from several traditional methods using a single classifier, our proposed ARPF needs more than four different initial classification maps acquired from different classifiers or image features. Second, an adaptive region around each pixel of the gray image is generated with two predefined parameters, and each adaptive region is applied to refine the corresponding pixel of each initial classified map. Finally, all the refined classified maps are merged to obtain the final classification map by coupling adaptive region and majority voting rules. In our experiments, three optical images with VHSR are used to evaluate the proposed ARPF. Compared with three typical relevant post-classification methods, the proposed ARPF can provide a classification map with less noise in visual performance and achieve higher quantitative accuracy while having an advantage in the constant detail of ground targets.

Learning from multimodal and multitemporal earth observation data for building damage mapping

Earth observation (EO) technologies, such as optical imaging and synthetic aperture radar (SAR), provide excellent means to continuously monitor ever-growing urban environments. Notably, in the case of large-scale disasters (e.g., tsunamis and earthquakes), in which a response is highly time-critical, images from both data modalities can complement each other to accurately convey the full damage condition in the disaster aftermath. However, due to several factors, such as weather and satellite coverage, which data modality will be the first available for rapid disaster response efforts is often uncertain. Hence, novel methodologies that can utilize all accessible EO datasets are essential for disaster management. In this study, we developed a global multimodal and multitemporal dataset for building damage mapping. We included building damage characteristics from three disaster types, namely, earthquakes, tsunamis, and typhoons, and considered three building damage categories. The global dataset contains high-resolution (HR) optical imagery and high-to-moderate-resolution SAR data acquired before and after each disaster. Using this comprehensive dataset, we analyzed five data modality scenarios for damage mapping: single-mode (optical and SAR datasets), cross-modal (pre-disaster optical and post-disaster SAR datasets), and mode fusion scenarios. We defined a damage mapping framework for semantic segmentation of damaged buildings based on a deep convolutional neural network (CNN) algorithm. We also compared our approach to another state-of-the-art model for damage mapping. The results indicated that our dataset, together with a deep learning network, enabled acceptable predictions for all the data modality scenarios. We also found that the results from cross-modal mapping were comparable to the results obtained from a fusion sensor and optical mode analysis.

Deep multisensor learning for missing-modality all-weather mapping

Multisensor Earth observation has significantly accelerated the development of multisensor collaborative remote sensing applications such as all-weather mapping using synthetic aperture radar (SAR) images and optical images. However, in the real-world application scenarios, not all data sources may be available, namely, the missing-modality problem, e.g., the poor imaging conditions obstruct the optical sensors, and only SAR images are available for the mapping. This real-world scenario raises the challenge of how to leverage historical multisensor data to improve the representation ability of the available model. The knowledge transfer based and knowledge distillation based approaches, as feasible solutions, can be used to transfer knowledge from other sensor models to the available model. However, these approaches suffer from the problem of forgotten knowledge and the multi-modality co-registration problem, which means that the leveraging of historical multisensor data is inefficient. The essential problem lies in the fact that these approaches are designed following the single-sensor data-driven approach. In this paper, a registration-free multisensor data-driven learning method, namely, deep multisensor learning, in a new perspective of knowledge retention, is proposed to overcome the above problems by learning a meta-sensory representation. To explore the existence of the meta-sensory representation, the meta-sensory representation hypothesis is first proposed, which reveals that the essential difference of the deep models trained on data from different sensors lies in the parameter distribution of the sensor-invariant and sensor-specific operations. Based on this hypothesis, a prototype network is proposed to learn the meta-sensory representation by modeling the knowledge retention mechanism, using the proposed difference alignment operation (DiffAlignOp). DiffAlignOp enables the prototype network to dynamically generate sensor-specific networks to gather supervised signals from registration-free multisensor data. This dynamic network generation is differentiable. Therefore, multisensor gradients can be obtained to learn the meta-sensory representation. To demonstrate the flexibility and practicality of deep multisensor learning, the application of all-weather mapping in a missing-modality scenario was performed. The experiments were conducted on a large public multisensor all-weather mapping dataset, which consists of high-resolution optical and SAR imagery with a spatial resolution of 0.5 m. The experimental results suggest that deep multisensor learning is superior to the other learning approaches in performance and stability, and reveals the importance of meta-sensory representation in multisensor remote sensing applications.

Inventory and changes of rock glacier creep speeds in Ile Alatau and Kungöy Ala-Too, northern Tien Shan, since the 1950s

Abstract. Spatio-temporal patterns related to the viscous creep in perennially frozen sediments of rock glaciers in cold mountains have rarely been studied outside the densely populated European Alps. This study investigates the spatial and temporal variability of rock glacier movement in the Ile Alatau and Kungöy Ala-Too mountain ranges, northern Tien Shan, a region with particularly large and fast rock glaciers. Over the study region of more than 3000 km2, an inventory of slope movements was constructed using a large number of radar interferograms and high-resolution optical imagery. The inventory includes more than 900 landforms, of which around 550 were interpreted as rock glaciers. Out of the active rock glaciers inventoried, 45 are characterized by a rate of motion exceeding 100 cm/a. From these fast rock glaciers we selected six (Gorodetzky, Morenny, Archaly, Ordzhonikidze, Karakoram, and Kugalan Tash) and studied them in more detail using offset tracking between repeat aerial images and historical and modern high-resolution optical satellite data. Two of these rock glaciers showed a steady increase in decadal surface velocities from the 1950s onwards, with speeds being roughly 2 to 4 times higher in recent years compared to the 1950s and 1960s. Three rock glaciers showed similar accelerations over the last 1 to 2 decades but also phases of increased speeds in the 1960s. This development indicates a possible significant increase in current sediment and ice fluxes through rock glaciers and implies that their material transport in the region might gain geomorphodynamic importance relative to material transport by glaciers, assuming the latter decreases together with the regional glacier shrinkage. The study demonstrates how air and satellite image archives are exploited to construct one of the longest decennial times series of rock glacier speeds currently available. Our results are in line with findings from Europe about rock glacier speeds increasing with atmospheric warming and underline local variability of such an overall response.

Decadal-scale hotspot methane ebullition within lakes following abrupt permafrost thaw

Thermokarst lakes accelerate deep permafrost thaw and the mobilization of previously frozen soil organic carbon. This leads to microbial decomposition and large releases of carbon dioxide (CO2) and methane (CH4) that enhance climate warming. However, the time scale of permafrost-carbon emissions following thaw is not well known but is important for understanding how abrupt permafrost thaw impacts climate feedback. We combined field measurements and radiocarbon dating of CH4 ebullition with (a) an assessment of lake area changes delineated from high-resolution (1–2.5 m) optical imagery and (b) geophysical measurements of thaw bulbs (taliks) to determine the spatiotemporal dynamics of hotspot-seep CH4 ebullition in interior Alaska thermokarst lakes. Hotspot seeps are characterized as point-sources of high ebullition that release 14C-depleted CH4 from deep (up to tens of meters) within lake thaw bulbs year-round. Thermokarst lakes, initiated by a variety of factors, doubled in number and increased 37.5% in area from 1949 to 2009 as climate warmed. Approximately 80% of contemporary CH4 hotspot seeps were associated with this recent thermokarst activity, occurring where 60 years of abrupt thaw took place as a result of new and expanded lake areas. Hotspot occurrence diminished with distance from thermokarst lake margins. We attribute older 14C ages of CH4 released from hotspot seeps in older, expanding thermokarst lakes (14CCH4 20 079 ± 1227 years BP, mean ± standard error (s.e.m.) years) to deeper taliks (thaw bulbs) compared to younger 14CCH4 in new lakes (14CCH4 8526 ± 741 years BP) with shallower taliks. We find that smaller, non-hotspot ebullition seeps have younger 14C ages (expanding lakes 7473 ± 1762 years; new lakes 4742 ± 803 years) and that their emissions span a larger historic range. These observations provide a first-order constraint on the magnitude and decadal-scale duration of CH4-hotspot seep emissions following formation of thermokarst lakes as climate warms.

Self-sufficiency through urban agriculture: Nice idea or plausible reality?

With more than half of the world’s population residing in cities, there has been intense interest in urban agriculture’s contribution to sustainability and self-sufficiency. This paper provides new insights into the potential for food production to occur within residential properties. High-resolution optical imagery was employed to quantify residential land immediately available for vegetable production (i.e. residential lawn) in a low-density city. Adelaide (Australia) was used as a case study. Productivity was estimated according to three empirical (low, medium and high) and one commercial yield scenario. Under high and medium yield scenarios, lawn in a typical block exceeded the space required to achieve household self-sufficiency of vegetables. Under the high-yield scenario, 23 % of lawn area would be required to meet the recommended vegetable intake of residents, while under the medium yield scenario, 72 % would be required. This study demonstrates the plausible reality of self-sufficiency through backyard vegetable production. Other resources that might limit production, such as labour, water and nutrients, are explored in the discussion. Urban agriculture is shown to be particularly worthwhile in low density cities; contrasting previous research performed under different urban conditions. However, in this case study, potentially productive land has decreased over time.

An Object-Based Approach for Mapping Tundra Ice-Wedge Polygon Troughs from Very High Spatial Resolution Optical Satellite Imagery

Very high spatial resolution commercial satellite imagery can inform observation, mapping, and documentation of micro-topographic transitions across large tundra regions. The bridging of fine-scale field studies with pan-Arctic system assessments has until now been constrained by a lack of overlap in spatial resolution and geographical coverage. This likely introduced biases in climate impacts on, and feedback from the Arctic region to the global climate system. The central objective of this exploratory study is to develop an object-based image analysis workflow to automatically extract ice-wedge polygon troughs from very high spatial resolution commercial satellite imagery. We employed a systematic experiment to understand the degree of interoperability of knowledge-based workflows across distinct tundra vegetation units—sedge tundra and tussock tundra—focusing on the same semantic class. In our multi-scale trough modelling workflow, we coupled mathematical morphological filtering with a segmentation process to enhance the quality of image object candidates and classification accuracies. Employment of the master ruleset on sedge tundra reported classification accuracies of correctness of 0.99, completeness of 0.87, and F1 score of 0.92. When the master ruleset was applied to tussock tundra without any adaptations, classification accuracies remained promising while reporting correctness of 0.87, completeness of 0.77, and an F1 score of 0.81. Overall, results suggest that the object-based image analysis-based trough modelling workflow exhibits substantial interoperability across the terrain while producing promising classification accuracies. From an Arctic earth science perspective, the mapped troughs combined with the ArcticDEM can allow hydrological assessments of lateral connectivity of the rapidly changing Arctic tundra landscape, and repeated mapping can allow us to track fine-scale changes across large regions and that has potentially major implications on larger riverine systems.

Fusion of High-Resolution SAR and Optical Imageries Based on a Wavelet Transform and IHS Integrated Algorithm

Multi-sensor remote sensing data can significantly improve the interpretation and usage of large volume data sources. A combination of satellite Synthetic Aperture Radar (SAR) data and optical sensors enables the use of complementary features of the same image. In this paper, SAR data is injected into optical image using a combining fusion method based on the integration of wavelet Transform and IHS (Intensity, Hue, and Saturation) transform. Not only to preserve the spectral information of the original (MS) image, but also to maintain the spatial content of the high-resolution SAR image. Two data sets are used to evaluate the proposed fusion algorithm: one of them is Pleiades, Turkey and the other one is Boulder, Colorado, USA. The different fused outputs are compared using different image quality indices. Visual and statistical assessment of the fused outputs displays that the proposed approach has an effective translation from SAR to the optical image. Hence, enhances the SAR image interpretability.

Landslide mapping using object-based image analysis and open source tools

Abstract Availability of high-resolution optical imagery and advances in image processing technologies have significantly improved our ability to map landslides. In recent years object-based image analysis (OBIA) has been gaining in popularity for landslide mapping due to its ability to incorporate spectral, textural, morphological and topographical properties. Many studies have been conducted based on commercial software. In this study, we create an open source Semi-Automatic Landslide Detection (SALaD) system utilizing OBIA and machine learning. Configured to run in Linux environment, it uses various open source Python packages and modules. This system was tested in 575 km2 area along the Pasang Lhamu Highway, Nepal where large numbers of landslides were triggered by the 2015 Gorkha earthquake. Comparison with a manual inventory highlighted that this system was able to detect 70% of the landslide area. The speed and efficiency with which this system was able to detect landslides makes it a viable alternative to manual techniques for landslide mapping over large areas, when establishing approximate landslide locations is of prime importance.

Arbitrary-Oriented Ship Detection via Feature Fusion and Visual Attention for High-Resolution Optical Remote Sensing Imagery

ABSTRACT Ship detection for high-resolution optical remote sensing (HRORS) imagery is widely applicable to Navy construction, maritime transportation, maritime safety, port management, and many other fields. Optical remote-sensing images can easily be disturbed by thin cloud, light, and some complex environmental conditions, which makes the task of ship detection challenging. In this paper, we propose a new general detection network framework to recognize ships with arbitrary-orientation, in which we use a novel way to define the rotating rectangular box, remove the dependence on methods with angle parameters of the target. We propose a new feature fusion module (FFM) to integrate the semantic information, which enhances the feature extraction from the shallow network, and effectively improve the detection of small targets. We also employ one attention module (AM) to adaptively select meaningful features, which enhance the ability of feature extraction for complex environmental conditions. The multi-directional anchor size and length-width ratio are designed based on the clustering technique to improve the convergence speed and accuracy. Experiments on the public HRORS ship detection dataset HRSC2016 and the ship dataset SSD2020 collected from Google Earth have shown better performance compared with some conventional methods, which prove the feasibility and effectiveness of our proposed network architecture.

An Approach for Evaluating Multisite Radiometry Calibration of Sentinel-2B/MSI Using RadCalNet Sites

The Sentinel-2B offers high spatial resolution optical imagery from 13 bands in the visible, near infrared, and short-wave infrared range of the electromagnetic spectrum, providing enhanced continuity to monitoring global terrestrial surfaces and coastal waters. In this article, its radiometry calibration is evaluated using 37 clear-sky observations in 2018 from the four sites of radiometric calibration network for assuring data quality. However, since the single calibration results acquired under different surface and atmospheric conditions have different biases and different uncertainties, it is difficult to determine which calibration sample is much more trustable. In view of this, by assuming that the calibration samples are independent of each other, a state-of-the-art reference value is derived by combining 37 calibration samples using a weighted average method, which has much lower uncertainty and approaches the “true” value. The reference value also could be used to compare each calibration result. The result shows that the reference value of the relative difference between the simulated and observed top-of-atmosphere reflectance is 4.29%, 4.95%, 4.54%, 5.34%, respectively, for bands 2, 3, 4, and 8, and the corresponding uncertainty is 1.09%, 1.10%, 1.10%, and 1.12%, respectively; the degree of equivalence for each sample is calculated by comparing each calibration result with the reference value. It is worth noting that the degrees of equivalence are lower than 5%, and the four samples on July 9, July 13, October 4, and October 11 perform worse than the other sample.