Lightweight Robust Target Detection Network in Barrage Jamming Environment Based on Multiscale Feature Extraction

With the advancement of active jamming technology, radar anti-jamming has become a significant research focus. Traditional anti-barrage jamming (ABJ) methods rely heavily on prior knowledge, can counter only a limited number of jamming types, and suffer from high system complexity. To cope with these issues, this article proposes a feature mining-based ABJ method. This method employs deep convolutional neural networks (CNNs) to extract high-dimensional features of the echo signal, thereby reducing the impact of barrage jamming and improving radar target detection probability. First, the echo signal under barrage jamming conditions is transformed into a range-Doppler spectrum image. This image is then fed into the proposed target detection network, which efficiently extracts multiscale features, enabling robust target detection, ranging, and speed measurement. In addition, the use of a two-layer Focus module for downsampling and feature channel expansion of the input range-Doppler spectrum image makes the proposed network lightweight and more suitable for practical engineering applications. Simulation results indicate that the proposed method achieves significantly higher target detection probabilities across 16 barrage jamming scenarios compared to other methods.