Deep CNN-Based Radar Detection for Real Maritime Target Under Different Sea States and Polarizations

Maritime target detection, because of the difficulties in the extraction and recognition of target and clutter micro-motion characteristics, has always been one of the difficulties in radar target detection. In this paper, convolutional neural networks are used for the detection of maritime target micro-Doppler. Firstly, using the IPIX measured sea clutter and target signal data, the two-dimensional time-frequency signal dataset is built by time-frequency analysis. Two Deep CNN models, LeNet and GoogLeNet are trained and used for the detection of maritime targets, and their performances are compared. Then the method is tested under different sea states and polarization. The results show that the proposed method can achieve high detection probability under different circumstance, which provides a new approach for the detection of maritime targets.