Adaptive CFAR Detection Based on Generalized Statistical Model

Constant False Alarm Rate (CFAR) processing, a critical automated target detection method in radar systems, has been extensively studied for its capabilities. Nonetheless, complex backgrounds and the advancements in stealth technology pose considerable challenges for CFAR in detecting dim targets. To tackle this issue, an adaptive CFAR detection algorithm is proposed that employs two statistically robust models, the Fisher distribution and generalized gamma distribution (GΓD) for clutter modeling. Leveraging the strengths of neural networks in online model recognition, this method can achieve CFAR detection of low signal-to-clutter ratio (SCR) signals in diverse clutter scenarios. Experimental results using high-resolution range profiles (HRRP) demonstrate that, in comparison to traditional CFAR detection methods, this approach exhibits superior performance.