Few-Shot Class-Incremental SAR Target Recognition with a Forward-Compatible Prototype Classifier

This tutorial presents a state-of-the-art survey of advanced applications of Synthetic Aperture Radar (SAR). Applications include: the detection and recognition of targets in SAR imagery, and the detection of targets under trees (and in the open) using change detection (coherent change detection, full-polarization change detection, etc). A fully polarimetric SAR sensor is described. Models of clutter and targets are developed from analysis of full-polarization SAR imagery. Algorithms for optimum (GLRT) processing of full-polarization data are derived, including the polarimetric whitening filter (PWF) and the polarimetric matched filter (PMF). Polarization-based discrimination features (polarization entropy, anisotropy, etc.) are developed and applied to SAR image terrain classification. Model-based and template-based target recognition systems are described and classifier performance results versus resolution and polarization are summarized for 10-and 20-target classifiers. Super-resolution algorithms are applied to SAR imagery and target recognition performance improvements demonstrated. Multi-polarization and multi-pass change detection algorithms are developed and demonstrated using UHF SAR imagery of targets under trees; fusion of multi-polarization coherent and non-coherent change detection algorithms is demonstrated. The effect of SAR image compression on change detection performance is quantified. Change detection performance using multi-pass VHF SAR imagery of targets under trees is demonstrated using data gathered by the Swedish CARABAS SAR system.

Recognizing the category attributes from the real world targets is one of the most challenging and attractive fields in synthetic aperture radar (SAR) application. It is an important issue to explore the spatial distribution characteristics of multi-aspect angle imagery in synthetic aperture radar automatic target recognition (SAR ATR). In this paper, we will research the spatial structure of multi-aspect angle SAR imagery through a visualization approach with real SAR data. Based on nonlinear dimensionality reduction, the representation of SAR samples is revealed in the low-dimensional Euclidean space, and the the nonlinear manifold distribution of multi-aspect angle SAR imagery is discovered. Besides, the regularity of that spatial distribution is summarized, i.e. the intrinsic structure of SAR images is parameterized by the aspect angles. The results of our research can provide a theoretical basis for SAR image classification and recognition algorithm designing.

In this paper, a novel moving target detection and tracking approach is presented based on Gaussian mixture probability hypothesis density (GMPHD) filter in synthetic aperture radar (SAR). Based on the advantages of GMPHD filter and the characteristics of moving target in SAR imagery, GMPHD filter is employed on potential moving target candidates extracted from a sequence of temporal and spatial sub-aperture SAR images to detect and track the moving targets in heavy clutter environment. Utilizing the tracking algorithm, the states and the number of moving targets are obtained over time. Simultaneously, the strong ground stationary clutter lacking of the dynamic behavior is eliminated finally. Both simulation and real Gotcha data set processing results are provided to demonstrate the effectiveness of the proposed approach.

This thesis describes the application of the Cramer Rao inequality to synthetic aperture radar (SAR). The resulting Cramer Rao lower bounds (CRB) reveal the smallest possible error variances for target parameter estimates (position and reflectivity), regardless of the estimator or SAR processing. The performance bounds are consequently used to design synthetic arrays and evaluate multistatic SAR configurations. We show that optimal sensor arrangements of synthetic arrays offer the possibility for single-pass, monostatic SAR to mitigate an effect known as layover that often distorts radar imagery. We also show these nonlinear apertures can be used to image targets in three dimensions. Furthermore, a new detection/estimation algorithm for targets in SAR imagery is developed and mean-squared errors from Monte Carlo simulations are compared to corresponding CRBs. Finally, a dynamic model-based estimation algorithm is developed for SAR to localize targets behind building walls. This iterative, optimization technique shows the potential to avert the combinatorial complexity and local maximization associated with many classification problems requiring model-based solutions.

Ship detection in synthetic aperture radar (SAR) images is becoming a research hotspot. In recent years, as the rise of artificial intelligence, deep learning has almost dominated SAR ship detection community for its higher accuracy, faster speed, less human intervention, etc. However, today, there is still a lack of a reliable deep learning SAR ship detection dataset that can meet the practical migration application of ship detection in large-scene space-borne SAR images. Thus, to solve this problem, this paper releases a Large-Scale SAR Ship Detection Dataset-v1.0 (LS-SSDD-v1.0) from Sentinel-1, for small ship detection under large-scale backgrounds. LS-SSDD-v1.0 contains 15 large-scale SAR images whose ground truths are correctly labeled by SAR experts by drawing support from the Automatic Identification System (AIS) and Google Earth. To facilitate network training, the large-scale images are directly cut into 9000 sub-images without bells and whistles, providing convenience for subsequent detection result presentation in large-scale SAR images. Notably, LS-SSDD-v1.0 has five advantages: (1) large-scale backgrounds, (2) small ship detection, (3) abundant pure backgrounds, (4) fully automatic detection flow, and (5) numerous and standardized research baselines. Last but not least, combined with the advantage of abundant pure backgrounds, we also propose a Pure Background Hybrid Training mechanism (PBHT-mechanism) to suppress false alarms of land in large-scale SAR images. Experimental results of ablation study can verify the effectiveness of the PBHT-mechanism. LS-SSDD-v1.0 can inspire related scholars to make extensive research into SAR ship detection methods with engineering application value, which is conducive to the progress of SAR intelligent interpretation technology.

One of the most straightforward techniques for detecting changes in an image involves forming the difference between a test image and a reference image. Unfortunately, such a technique can give rise to a large number of false alarms due to the statistical variability of the underlying pixel values, as has been well established within the radar community over the years. One method for dealing with this large number of false alarms involves forming the ratio, rather than the difference, of two synthetic aperture radar (SAR) images. We introduce a modified version of the standard differencing technique to overcome problems associated with pixel value variability. The new (modified) differencing approach utilizes assumptions about the statistics of the image background and the object being sought (target) to reduce the number of false alarms due to highly variable background (clutter) regions, and it includes the standard ratio test as a special case. In fact, we find that the modified difference approach can also be viewed as a modified version of the ratio test with a threshold that varies as a function of the background clutter radar cross section (RCS). We also present an abridged, albeit suboptimal, version of this approach that eliminates assumptions regarding the target's probability distribution, and we analyze both of the approaches. We then compare these results with those obtained with a standard ratio test, and illustrate how the modified difference test reduces to the ratio test under certain operating conditions. The abridged version of the modified approach is applied to high resolution synthetic aperture radar imagery and compared with results obtained with the classical differencing technique, and following this, the modified difference technique is compared with the standard ratio test. Results suggest that under appropriate conditions the abridged, modified technique can successfully detect changes without the need for any image segmentation.

Synthetic Aperture Radar (SAR) imagery is available all the time and, in all weathers, due to its ability to penetrate through cloud cover, dust, haze except heavy rainfall. SAR imagery mainly characterizes structural properties of various earth objects and provides a data which is rich in spatial information. Processing of SAR data has been of great interest due to these advantages. However, due to coherent processing of SAR signal, imagery is corrupted by speckle noise which is multiplicative in nature. Reduction of speckle noise from SAR imagery is helpful for analysis and interpretation of SAR images. In this paper, performance of conventional speckle reduction filters, namely Lee, Enhanced Lee, Frost, Enhanced Frost and Gamma MAP are compared with wavelet-based filtering technique for RISAT-1 imagery. Results are compared visually and using quality metrics, Speckle Suppression Index (SSI), Equivalent Number of Looks (ENL) and Speckle Suppression and Mean Preservation Index (SMPI). From visual interpretation of de-speckled image, it is evident that wavelet-based filtering out performs the conventional filters in terms of edge and texture preservation.

Oil spills on the ocean from ships have a strong impact on the environment as well as on human societies. The MV X-Press Pearl container ship incident in 2021 became the worst marine environmental disaster in Sri Lanka due to its powerful impact. Oil leakage from the ship to the ocean continued for months. Synthetic Aperture Radar (SAR) imagery has the advantage of penetrating cloud cover, as well as day and night operating capabilities. Therefore, SAR imagery can be used to detect as well as observe the temporal changes of oil spills. The study in this article aimed to detect the oil spill caused by the MV X-Press Pearl container ship on Sri Lanka’s west coast from May to August 2021 using Sentinel-1 SAR data and measure the temporal changes of the oil slick. Three SAR imageries with Vertical-Vertical (VV) and Vertical-Horizontal (VH) polarization from June, July, and August of 2021 have been processed by applying the orbit file, running the multi-looking algorithm, and making the ellipsoid correction. The possible oil spills have been identified from each set of imagery as dark linear shapes with lengths of 3.1 ±0.012 km with an average reflectance of 15.6 decibels for June and 3.6 ±0.012 km with an average reflectance of 16.5 decibels for the other two months, and the shapes correspond with the oil slick classifications from other literature. The results indicate that the VV polarization is the most suitable to detect oil spills, and that the oil spill from the X-Press Pearl continued from June to the end of August 2021.

Normalized Difference Vegetation Index (NDVI) values extracted from remotely sensed optical imagery are used ubiquitously to monitor crop condition. However, challenges in the operational use of optical imagery are well documented making it difficult to capture measures of crop condition during critical phenology stages when clouds obscure. This study investigates the integration of Synthetic Aperture Radar (SAR) and optical imagery to characterize the condition of crop canopies in order to deliver daily measures of NDVI during the entire growing season. Multitemporal C-band polarimetric RADARSAT-2 SAR data and RapidEye images were acquired in 2012 for a study site in western Canada. SAR polarimetric parameters and NDVI were extracted. The temporal variations in SAR polarimetric parameters and NDVI were interpreted with respect to the development of the canola canopy. Optical NDVI was statistically related with SAR polarimetric parameters over test canola fields. Significant correlations were documented between a number of SAR polarimetric parameters and optical NDVI, in particular with respect to HV backscatter, span, volume scattering of the Freeman Durden decomposition and the radar vegetation index, with R-values of 0.83, 0.72, 0.81 and 0.71 respectively. Based on the statistical analysis, SAR polarimetric parameters were calibrated to optical NDVI, creating a SAR-calibrated NDVI (SARc-NDVI)). A canopy structure dynamics model (CSDM) was fitted to the SARc-NDVI, providing a seasonal temporal vegetation index curve. The coupling of NDVI from optical and SAR imagery with a CSDM demonstrates the potential to derive daily measures of crop condition over the entire growing season.

Synthetic Aperture Radar (SAR) imagery have been one of the important tools to support earth observations and topographic measurements. It means SAR imagery are essentially rich in structures and some important target categories are difficult to recognize. Optical imagery contains rich and clear spectral information which has a good influence on semantic image segmentation. The success of deep neural networks for semantic segmentation heavily depends on large-scale and well-labeled data sets, which are hard to collect in practice. In this paper, we consider deep transfer learning for semantic segmentation, we propose a deep novel transfer learning method, which transfers a semantically segment model from SAR imagery to SAR and optical fusion imagery. The experimental results show that the method proposed achieves higher mean Intersection over Union (mIoU) with less training time compared with other methods.

A generalized control chart for anomaly detection in SAR imagery

Although deep learning techniques have achieved noticeable success in aircraft detection, the scale heterogeneity, position difference, complex background interference, and speckle noise keep aircraft detection in large-scale synthetic aperture radar (SAR) images challenging. To solve these problems, we propose the geospatial transformer framework and implement it as a three-step target detection neural network, namely, the image decomposition, the multiscale geospatial contextual attention network (MGCAN), and result recomposition. First, the given large-scale SAR image is decomposed into slices via sliding windows according to the image characteristics of the aircraft. Second, slices are input into the MGCAN network for feature extraction, and the cluster distance nonmaximum suppression (CD-NMS) is utilized to determine the bounding boxes of aircraft. Finally, the detection results are produced via recomposition. Two innovative geospatial attention modules are proposed within MGCAN, namely, the efficient pyramid convolution attention fusion (EPCAF) module and the parallel residual spatial attention (PRSA) module, to extract multiscale features of the aircraft and suppress background noise. In the experiment, four large-scale SAR images with 1-m resolution from the Gaofen-3 system are tested, which are not included in the dataset. The results indicate that the detection performance of our geospatial transformer is better than Faster R-CNN, SSD, Efficientdet-D0, and YOLOV5s. The geospatial transformer integrates deep learning with SAR target characteristics to fully capture the multiscale contextual information and geospatial information of aircraft, effectively reduces complex background interference, and tackles the position difference of targets. It greatly improves the detection performance of aircraft and offers an effective approach to merge SAR domain knowledge with deep learning techniques.

This paper proposes using a backpropagation (BP) neural network for the classification of ship targets in airborne synthetic aperture radar (SAR) imagery. The ship targets consisted of 2 destroyers, 2 cruisers, 2 aircraft carriers, a frigate and a supply ship. A SAR image simulator was employed to generate a training set, a validation set, and a test set for the BP classifier. The features required for classification were extracted from the SAR imagery using three different methods. The first method used a reduced resolution version of the whole SAR image as input to the BP classifier using simple averaging. The other two methods used the SAR image range profile either before or after a local-statistics noise filtering algorithm for speckle reduction. Performance on an extensive test set demonstrated the performance and computational advantages of applying the neural classification approach to targets in airborne SAR imagery. Improvements due to the use of multi-resolution features were also observed.

Data fusion technique is an efficient way to benefit multi-source, multi-platform, and multi-angle remotely sensed information. Optical imagery and SAR (synthetic aperture radar) data are complementary in terms of capability of data acquisition and image characteristics. With their different capability and their unique information content respectively, fusion of high resolution SAR and optical multi-spectral imagery can improve the classification accuracy in land use. Texture information plays an important role for class discrimination especially in SAR imagery for its backscatter is sensitive to the type, orientation, homogeneity and spatial relationship of ground objects. In order to take full advantage of multi-source remotely sensed data and combine different features of them, this paper put forward a data fusion method for high spatial resolution remotely sensed data based on texture analysis. Texture features of high resolution SAR imagery were extracted using GLCM (Grey Level Co-occurrence Matrix) method. The texture features were detected in 0°, 45°, 90° and 135° four directions, and the moving window size of 3×3, 5×5, to 31×31, 41×41, 51×51, and 61×61 were tested to analyze the influences among them. The selected texture features were added with SAR data to make classification next. Both the two imagery were classified using an object-based and rule-based approach. Then, a decision level fusion was implemented and the accuracy of classification result was improved from 78.7% and 83.0% to 88.8%.

Monitoring glacier change processes leads to a better understanding of how glaciers respond to various external forcings. For sub-annual remote sensing, Synthetic Aperture Radar (SAR) imagery is indispensable, as polar night and cloud cover limit the temporal continuity of optical imagery. Especially with the launch of the Sentinel-1 mission, the availability of suitable SAR imagery has increased substantially. This high amount of data leads to another challenge: Manual inspection is no longer feasible. Therefore, in recent years, many studies applied deep learning to automate the segmentation of calving fronts in satellite imagery. To ensure comparability between different deep learning models, they must be trained and tested on the same data with a predefined test set and evaluated with the same metrics. A dataset intended to be used for this purpose is called a benchmark dataset and needs to provide the ready-to-use satellite imagery and the corresponding calving front labels. Gourmelon et al. [1] provide such a dataset for SAR imagery of calving fronts called CaFFe (https://doi.pangaea.de/10.1594/PANGAEA.940950). CaFFe includes multi-mission data (ERS-1/2, RADARSAT 1, Envisat, ALOS, Sentinel-1A/B, TerraSAR-X, and TanDEM-X), providing a spatial resolution between 6 and 20 meters and covering the period from 1996 to 2020. It contains images of seven glaciers from Antarctica to Greenland and Alaska. For each of the 681 images contained in the benchmark, two labels are provided: One displaying the calving front versus background and the other showing different zones (ocean, rock outcrops, glacier area, and no information available). CaFFe is split into a train set and a predefined out-of-sample test set, which comprises all images from two of the seven glaciers. A split of the train set into training and validation is not specified, as different approaches like cross-validation shall be possible. The benchmark covers a wide variety of different conditions to capture the variability of SAR calving front images. For example, images with open oceans and images with ice-melange-covered oceans are included in the dataset. Especially including images featuring ice-melange is of great importance, as deep learning models have shown difficulties in accurately segmenting calving fronts under this condition. Including images with ice-melange in the train set helps models to learn accurate predictions even under these circumstances.  Adding such images to the test set ensures that evaluated models are able to cope with ice melange. The test set of CaFFe is specifically designed to be challenging, such that the generalizability of models to different conditions and spatial transferability even to other continents can be verified. Gourmelon et al. provide baselines (one for each of the available labels) complementing the benchmark dataset. Current collaborative work aims to evaluate recently published deep learning techniques for calving front extraction on CaFFe and compare it with the baselines.[1] N. Gourmelon, T. Seehaus, M. Braun, A. Maier, and V. Christlein: "Calving Fronts and Where to Find Them: A Benchmark Dataset and Methodology for Automatic Glacier Calving Front Extraction from SAR Imagery," Earth System Science Data, vol. 14, no. 9, pp. 4287-4313, 2022, doi: 10.5194/essd-14-4287-2022.