Ship Detection for SAR Images Based on Fusion of Superpixel Segmentation and Aggregation

A method of ship detection for SAR images based on fusion of superpixel segmentation and aggregation is proposed in this paper. First, two new clustering center shift criterion and adaptive similarity measurement criterion are defined, and SAR image superpixels are generated based on these two criteria. Then, the texture similarity between the superpixels is calculated by the multidirectional gray-level co-occurrence matrix, and the superpixels with similar textures are aggregated. Finally, based on the results of superpixel aggregation, the rule-based target detection method is used to realize the rapid ship detection. The experimental results based on various types of SAR images show that the proposed method has robust ship detection performance.