Periodicity Detection: Recognition Between Target and Deception Jamming in Radar Systems via HAM-YOLOv8n Network

Recognizing physical target (PT) from advanced false target (FT) with similar electromagnetic characteristics is a critical challenge for radar systems. While artificial intelligence (AI) offers potential solutions, existing methods are trained only on data from specific scenarios and ignore the dynamic nature of detection environments (including targets and jamming), resulting in a lack of generalization capability in dynamic operational scenarios. This paper proposes a periodicity detection-based recognition method within a Netted Radar (NR) framework to address this limitation. The method leverages the fundamental differences in scattering characteristics between PT and FT from a NR perspective (i.e., the diversity of PT spectrum features versus the consistency of FT spectrum features). Through coherent integration (CI), the method transfers the discriminative spectrum features to the fast-slow time domain and constructs the NR's fast-slow time domain amplitude spectrum (NRFST-AS) feature maps for PT and FT. Then, a lightweight hybrid attention mechanism-YOLOv8n (HAM-YOLOv8n) network is designed to efficiently and accurately detect the periodic and non-periodic patterns of targets and jamming in the feature maps, thereby accomplishing recognition. Experimental results from both simulations and measured data demonstrate that the proposed method not only performs stably under low signal-to-noise ratio (SNR) conditions but also exhibits excellent generalization ability for untrained scattering types and scenarios. The recognition accuracy on measured data exceeds 96.5$\%$, which is significantly better than that of comparison methods.