Small Target Detection in Sea Clutter Based on SDP Image Features

The nonstationary, non-Gaussian, and nonhomogeneous sea clutter pose significant challenges for the detection of small floating targets in sea clutter. Traditional detectors based on statistical models often encounter performance limitations. However, recent research has shown that echo signals contain abundant frequency, amplitude, and phase information, which is of great significance for the development of feature-based detector. Due to the ordered target and the chaotic clutter, we extract the symmetric dot pattern (SDP) image features to capture the regularity of signal in the time dynamic trends and propose a threshold controllable random forest detector (SDP-TC-RF) to detect the small target in sea clutter. The amplitude of signal is transformed by SDP into mirror-symmetric snowflake images in a 2-D polar coordinate space, referred to as SDP images. It is demonstrated that target SDP images exhibit an ordered structure, such as sharp boundary definition, morphological smoothness, and rotational symmetry, while clutter SDP images appear disordered, resembling the distinction between nonchaotic targets and chaotic clutter. Subsequently, texture features and edge features of both the sea clutter SDP images and the target SDP images are extracted and fed into the random forest (RF) detector with a control the false alarm rate based on the Neyman–Pearson criterion. To leverage the different features extracted from the SDP images, we design two detectors: the TC-RF detector based on the texture features of the SDP images (SDP-T-TC-RF) and the TC-RF detector based on the Sobel edge features of the SDP images (SDP-S-TC-RF). Experimental validation is carried out with the publicly available IPIX radar dataset. With an accumulation time of 1.024 s and a false alarm probability of 0.001, the average detection probabilities of the SDP-T-TC-RF and SDP-S-TC-RF detectors reach 0.5852 and 0.9316, respectively, demonstrating an effective detection performance.