Weak Target Detection in Radar Sea Clutter Based on Weighted Convex Hull Tree Algorithm

Detecting small targets on the sea with high-resolution radars is difficult because of weak reflections and complex sea clutter. Using multiple features to distinguish targets from clutter is an effective method. Thus, creating a specialized classifier that can handle unequal training samples, especially ergodic sea clutter versus non-ergodic target returns, is crucial. This paper proposes a weighted convex hull tree structure with three parts: convex hull, weight, and threshold decision. First, features are combined to form multiple three-dimensional convex hulls. Second, different convex hulls are weighted based on their detection capabilities about simulated target samples. Finally, a test statistic is built using weights and sample distances to control false alarms. The algorithm uses measured sea clutter and simulated target data, ensuring robustness. On the open IPIX database and a self-collected UAV dataset, the proposed detector outperforms other feature-based detectors on the performance and computation cost.