A Two-Level Cascade Detection Algorithm for X-Band Marine Radar

X-Band marine radar plays an important role in maritime detection. The targets are the extended targets due to the high resolution and the strong reflection targets have strong sidelobe. Besides, there are much clutter and noises in the radar echo. In this letter, we propose a two-level cascade detection algorithm for X-Band marine radar. The first-level detector, which consists of CFAR detector, Ostu algorithm and opening operation, aims to detect the suspicious extended targets as many as possible. Firstly, the CFAR detector is utilized to suppress the clutter and sidelobe. Then the Ostu algorithm and opening operation are utilized to segment the extended targets. Furthermore, we impose different ratios for the targets to handle different targets scales. The designed lightweight convolutional neural network (CNN), which combines the AlexNet design idea and convolutional block attention module (CBAM), acts as the second-level detector to distinguish between the true targets and false alarms. Experiments show that the proposed two-level cascade detection algorithm has satisfying detection performance