A Shape-Context Descriptor-Based Approach for Key Area Identification in Ship Detection by Passive Interferometric Microwave Sensor

Passive microwave interferometric radiometer imaging has emerged as a powerful tool for ship detection, especially under conditions of low-visibility and high sea clutter. After ships are correctly detected, an important task is the accurate identification of key areas of ship targets. In this article, we propose an effective and robust key area recognition and localization approach based on the shape-context descriptor. After extracting the target contour using Chan–Vese (CV) model, we apply the ellipse fitting method for orientation correction and incorporate the shape-context descriptor to describe shape features. Then, we present similarity measurement matrix to localize key areas. Experiment results demonstrate the robustness of the shape-context module in distinguish the bow and stern areas, even with blurred target contours. The numerical results prove that the proposed method shows satisfactory recognition accuracy, and competitive localization precision.