Dimension Compressed CFAR for Massive MIMO Radar

Abstract

This letter presents a novel method used to find the CFAR threshold after coherently compressing one or more dimensions of the radar image, generally the ones obtained from phased array or MIMO diversity. The method provides a significantly large improvement in the numerical complexity of the processing chain at the cost of a slightly lower probability of detection for a fixed probability of false alarm. The method is particularly useful for real-time applications such as assisted and autonomous driving, where CFAR is applied to the Range-Doppler map to generate a point cloud before angular estimation is done. The method may also be extended for Earth observation or surveillance applications where real-time processing is required. The proposed method is intended as a pre-processing step to the CFAR detection, and is compatible to any CFAR algorithm that assumes Gaussian noise. The method implies finding a CFAR threshold calculated from low-SNR non-coherent integrated radar signal and analytically map it to the high-SNR coherent integrated radar signal for improved detection. The proposed method is first presented theoretically and then evaluated in both simulated and experimental environments.