A Novel Small Target Detection Method for Marine Radar
Tianchang Gu, Wei Xu, Xuecheng Hu · 2019
This paper proposes a sea-surface small target detection method for marine radar. Six features (the relative amplitude, the relative Doppler peak height , the relative entropy of the Doppler amplitude spectrum and three time-frequency features) extracted from the radar returns are combined into 6-D feature vectors. Then considering that getting a labeled set of instances of all possible type of targets is unrealistic, a PCA-based anomaly detector with adjustable false alarm rate is constructed to discriminate the target bins with the sea clutter bins in the feature space. Adaptive weighted reconstruction errors of feature vectors are calculated and ranked and then test results are given. Experiments based on the measured data of a high-resolution X-band radar show that the proposed method attains higher detection probability than several existing detectors.