Over the recent years, deep-learning technology has been widely used. However, in research based on Synthetic Aperture Radar (SAR) ship target detection, it is difficult to support the training of a deep-learning network model because of the difficulty in data acquisition and the small scale of the samples. This paper provides a SAR ship detection dataset with a high resolution and large-scale images. This dataset comprises 31 images from Gaofen-3 satellite SAR images, including harbors, islands, reefs, and the sea surface in different conditions. The backgrounds include various scenarios such as the near shore and open sea. We conducted experiments using both traditional detection algorithms and deep-learning algorithms and observed the densely connected end-to-end neural network to achieve the highest average precision of 88.1%. Based on the experiments and performance analysis, corresponding benchmarks are provided as a basis for further research on SAR ship detection using this dataset.

Synthetic Aperture Radar three-dimensional (SAR 3D) imaging technology can eliminate severe overlap in 2D images, and improve target recognition and 3D modeling capabilities, which have become an important trend in SAR development. After decades of development of SAR 3D imaging technology, many types of 3D imaging methods have been proposed. In this study, the history of SAR 3D imaging technology is systematically reviewed and the characteristics of existing SAR 3D imaging technology are analyzed. Given that the 3D information contained in SAR echo and images is not fully used by existing techniques, a new concept of SAR microwave vision 3D imaging has been proposed for the first time. This new concept is integrated with microwave scattering mechanism and image visual semantics to realize three-dimensional reconstruction, which form the theory and method of SAR microwave vision 3D imaging and can achieve high-efficiency and low-cost SAR 3D imaging. This study also analyzes the concept, goal and key scientific problems of SAR microwave vision 3D imaging and provides a preliminary solution, which will contribute in several ways to our understanding of SAR 3D imaging and provide the basis for further research.

Flying birds and Unmanned Aerial Vehicles (UAVs) are typical “low, slow, and small” targets with low observability. The need for effective monitoring and identification of these two targets has become urgent and must be solved to ensure the safety of air routes and urban areas. There are many types of flying birds and UAVs that are characterized by low flying heights, strong maneuverability, small radar cross-sectional areas, and complicated detection environments, which are posing great challenges in target detection worldwide. “Visible (high detection ability) and clear-cut (high recognition probability)” methods and technologies must be developed that can finely describe and recognize UAVs, flying birds, and “low-slow-small” targets. This paper reviews the recent progress in research on detection and recognition technologies for rotor UAVs and flying birds in complex scenes and discusses effective detection and recognition methods for the detection of birds and drones, including echo modeling and recognition of fretting characteristics, the enhancement and extraction of maneuvering features in ubiquitous observation mode, distributed multi-view features fusion, differences in motion trajectories, and intelligent classification via deep learning. Lastly, the problems of existing research approaches are summarized, and we consider the future development prospects of target detection and recognition technologies for flying birds and UAVs in complex scenarios.

Deep learning such as deep neural networks has revolutionized the computer vision area. Deep learning-based algorithms have surpassed conventional algorithms in terms of performance by a significant margin. This paper reviews our works in the application of deep convolutional neural networks to target recognition and terrain classification using the SAR image. A convolutional neural network is employed to automatically extract a hierarchic feature representation from the data, based on which the target recognition and terrain classification can be conducted. Experimental results on the MSTAR benchmark dataset reveal that deep convolutional network could achieve a state-of-the-art classification accuracy of 99% for the 10-class task. For a polarimetric SAR image classification, we propose complex-valued convolutional neural networks for complex SAR images. This algorithm achieved a state-of-the-art accuracy of 95% for the 15-class task on the Flevoland benchmark dataset.