Multiscale Feature Learning Based on Deep Pyramid Residual Shrinking Network for Radar Target Detection

Boyu Wang,Gongjian Zhou

DOI: https://doi.org/10.1109/taes.2024.3488675

IF: 3.491

2024-01-01

IEEE Transactions on Aerospace and Electronic Systems

Abstract:Range migration (RM) and Doppler frequency migration (DFM) are unavoidable issues in the coherent integration of maneuvering targets. Existing deep learning-based trajectory detection methods still exhibit unsatisfactory detection performance under weak radar echo conditions due to inadequate feature extraction capabilities, and they are opaquely interpretable in architecture as black boxes. In this paper, a deep pyramid residual shrinking network (DPRSNet) for radar target detection is proposed. First, the pyramid shrinking module is designed to fully mine the discriminative shallow spatial feature information covered in the multiscale space from low signal-noise ratio radar echoes. Different sizes of large convolution kernels are adopted to expand the receptive field. Subsequently, the residual module is introduced to capture the middle and deep spatial feature information. Dense shortcut connection architectures are employed to enhance the feature information fusion and reduce redundancy. Additionally, the visual analysis is conducted to understand the feature information extracted by the network more intuitively. A threshold-selected strategy for sliding detection anchor box is seamlessly integrated into DPRSNet, improving the efficiency of the anchor box when sliding along the range. Extensive simulation results demonstrate that the proposed method outperforms state-of-the-art network in both trajectory detection accuracy and detection probability, with improvements of 17.2% and 26.22% respectively, underscoring the robustness and superiority of the proposed method. Benefiting from the incorporation of visualized feature maps, the interpretability of the results is enhanced and better comprehensibility is achieved.