Multi-Terrain Radar HRRP Target Detection Using Contrastive Learning

As radar resolution improves, high-resolution radars have significantly enhanced target detection performance through the generation of high-resolution range profiles (HRRPs). However, when the range resolution is much smaller than the target, the target's energy is dispersed across multiple range cells, leading to poor detection performance with traditional energy-based detection algorithms. Furthermore, existing detection algorithms are highly sensitive to clutter characteristics and exhibit poor adaptability to varying scenarios. To address these issues, this paper proposes a Constant False Alarm Rate (CFAR) HRRP detection algorithm based on contrastive learning for multiple terrains. The algorithm improves the contrastive learning loss function to achieve well-clustered representations of clutter samples during training. Test samples are evaluated by calculating the Mahalanobis distance between them and the training samples, and compared against detection thresholds to accomplish target detection. Testing results show that the proposed algorithm achieves superior detection performance across multiple terrains. Furthermore, to reach a 90% detection probability, the proposed method requires a signal-to-clutter ratio (SCR) reduction of 1 to 4 dB compared to the competing methods.