PMPE-Net: A High-Precision End-to-End Model for Radar Antijamming Target Detection

When the radar is subjected to deception jamming, its target detection performance will significantly decrease. Conventional anti-jamming methods usually adopt a multi-stage processing flow, including jamming perception, recognition, suppression, target detection, which often leads to high system redundancy, accumulated stage errors, and limited detection capability. To overcome these limitations, we propose an anti-jamming target detection network based on Prior-Guided Modulation and Pulse-Compression Embedded (PMPE-net), which achieves end-to-end integrated processing from raw radar echoes to jamming type discrimination and target location output. Firstly, we design a dual-modal hybrid jamming perception architecture to deeply mine and fuse discriminative jamming and target feature from different modalities. Secondly, we propose a multi-scale adaptive feature extraction architecture, which utilizes jamming recognition results as prior information to dynamically modulate the subsequent feature extraction process, effectively decoupling the features of different jamming types and the target. Finally, we construct a physics-and-data co-driven target detection architecture by embedding pulse compression layer, which enhances model interpretability and improves detection accuracy. Simulation results demonstrate that PMPE-net achieves superior end-to-end detection performance and strong generalization across diverse jamming scenarios. This study provides an efficient framework for high-performance radar target detection in complex electromagnetic environments.