Design and implementation of real-time detection method for moving target using through-wall radar based on single-frame multi-channel registration

Through-wall radar (TWR) motion target detection can provide real-time information on moving targets in enclosed environments, playing an important role in urban counter-terrorism and stability maintenance. Current research mainly focuses on TWR motion target detection algorithms, most of which suffer from high computational complexity, making it difficult to meet the requirements of real-time applications. To address this issue, a real-time TWR motion target detection method based on single-frame multi-channel registration and its efficient implementation is proposed. First, a clutter suppression method based on single-frame multi-channel data frequency-domain phase compensation and cancellation is introduced, which effectively reduces the data waiting latency inherent in traditional multi-frame data cancellation approaches. Secondly, a real-time detection method for moving targets of TWR based on multi stream asynchronous parallel and thread hierarchy of graphics processing unit (GPU) is designed. The proposed method includes a GPU-based clutter suppression multi-stream asynchronous parallel strategy, a coarse-and fine-grained parallelization of the back-projection (BP) imaging algorithm, and a block-level inter-parallel and intra-block reduction parallel strategy for two-dimensional constant false-alarm rate (CFAR) detection. Both simulation and experimental results demonstrate that the proposed method can complete real-time detection of moving targets behind walls within 1.078 ms for single-frame data with eight channels and 8192 range samples. Compared with conventional central processing unit-based implementations, the proposed method achieves a 92-fold improvement in processing efficiency, confirming its effectiveness and robustness.