Using Range-Doppler Spectrum-Based Deep Learning Method to Detect Radar Target in Interrupted Sampling Repeater Jamming

Main-lobe interrupted sampling repeater jamming (ISRJ) poses a serious threat to the normal use of radar. The existing anti-ISRJ target detection methods based on deep learning have problems, such as a large number of training samples required, complex use steps, inability to achieve ranging and speed measurement at the same time, and poor detection probability under a very low signal-to-noise ratio (SNR) environment. To solve these problems, this article proposes an end-to-end anti-ISRJ target detection method based on the range-Doppler spectrum. First, a visual object detection network based on convolutional neural networks (CNNs) is designed and trained with fewer than 9000 samples. Focal loss is introduced to train the network. It can adaptively improve the loss weight of hard samples and, therefore, improve the performance of the network under a low SNR environment. After that, the range-Doppler spectrum corresponding to the radar echo signal is input into the trained network, and then, the detection, ranging, and speed measurement of the targets under ISRJ conditions can be realized. The simulation results show that the detection probability of the proposed method under various SNRs, signal-to-jamming ratios (SJRs), and false alarm rates is better than that of the compared method. The proposed method can achieve a detection probability of more than 80% under the condition of a false alarm rate of $1e-4$ , an SNR of −20 dB, and an SJR of not less than −20 dB.