A method of ship wake detection in sar images based on reconstruction features and anomaly detector

An anomaly-detection-based method is proposed to improve the performance of ship wake detection in synthetic aperture radar (SAR) images of different sea states. The dictionaries learned by sea clutter images are poor to sparsely reconstruct wake images, since the nature of the sea clutter and ship wake is different. Therefore, the image reconstruction errors based on the sea clutter dictionaries are introduced to separate ship wake and sea clutter. The proposed method transforms the wake detection into an anomaly detection problem in the image reconstruction error feature space. First, the dictionaries are learned by a large number of sea clutter samples. Second, taking the sea clutter image reconstruction errors as discriminatory feature, errors of the wake will be larger than the threshold and the detection threshold is obtained by anomaly detectors. Finally, the ship wake detection is made by comparing the reconstruction errors from the test image samples with the detection threshold. Experimental results demonstrated the proposed method can improve the ship wake detection probability compared with the exiting method.