A Novel Postprocessing Method for Ship Detection in SAR Images Using Greedy Strategies

This article proposes a novel post-processing method for ship detection in synthetic aperture radar (SAR) images, which addresses the issue of head to tail separation that cannot be avoided when using morphological compensation to process candidate regions generated by the constant false alarm detection (CFAR). This algorithm first constructs a connected domain for the candidate regions generated by CFAR, uses a greedy strategy to update the centroid and size of the connected domain, and finally unreasonable candidate connected domains are removed to eliminate high noise sea clutter and large volume non ship targets. Compared to the morphological compensation post-processing method, the greedy strategy the greedy strategy fully utilizes the prior knowledge of ship data in the post-processing of constant false alarm ship detection results, overcomes the disadvantage of complex morphological compensation methods that require fine-tuning parameters in post-processing, and completely solves the problem of head tail separation that cannot be completely solved, improving the accuracy of synthetic aperture radar ship detection. The effectiveness and robustness of our method have been demonstrated by the data from the domestically produced Gaofen 3 satellite and the data from the European Space Agency's TerraSAR-X satellite.