Constant false alarm rate (CFAR) is a useful method in adaptive radar detection when the background noise is unknown. Most traditional CFAR methods cannot obtain a good tradeoff between homogeneous backgrounds and nonhomogeneous backgrounds.
With machine learning achieving ground breaking success in many research fields, intelligent CFAR detectors are considered for improving radar target detection in different backgrounds.
Recently, WANG Leiou and his team from the Institute of Acoustics of the Chinese Academy of Sciences proposed an intelligent CFAR detector based on support vector machine (SVM).
After training a SVM by using the variability index (VI) statistic as a feature, the proposed detector (SVM-CFAR) can recognize current operational environment based on the classification results of the SVM, and have the intelligence to select the proper detector threshold adaptive to the current operational environment.
At the training stage, a VI module is employed as priori data to train a SVM module. Based on the VI of current reference window and the test cell, the SVM module outputs a decision function during practical testing stage. A CFAR module, which contains various thresholds corresponding to different operational environments, provides an adaptive threshold according to the decision function.
The unique aspect of the SVM-CFAR is that it utilizes VI as a feature to train SVM module and provides the threshold based on classification results of the SVM module.
The detection performance of this well trained SVM-CFAR is compared to six conventional detectors in different environments.
Simulation results show that SVM-CFAR provides a low loss performance in homogeneous backgrounds. In the environments with a single interfering target, the SVM-CFAR detector is not affected by the position of the interfering target. For the case of multiple interfering targets both in the leading window and the lagging window, the SVM-CFAR is robust without excessive detection degradation. As far as false alarm is concerned, the SVM-CFAR is robust in clutter edges.
WANG Leiou, WANG Donghui, HAO Chengpeng. Intelligent CFAR Detector Based on Support Vector Machine. IEEE ACCESS （VOL. 5, NO. 1, November 2017）. DOI: 10.1109/ACCESS.2017.2774262.
Institute of Acoustics, Chinese Academy of Sciences, 100190, Beijing, China