In radar or sonar signal processing, constant false-alarm rate (CFAR) detection plays an important role, for it can set a decision about presence or absence of a target, using the adaptive threshold techniques.
CFAR detectors set the threshold adaptively according to the background information to maintain a constant false-alarm rate. In addition, to estimate the background power level accurately, the normalization is needed to realize the outlier rejection of the background data.
The Rayleigh CFAR detectors presume that the background is homogeneous and Rayleigh distributed. Besides, the non-Rayleigh CFAR detectors presume that the background is non-Rayleigh distributed. The non-Rayleigh distributions include Rayleigh mixture distribution, K-distribution, Weibull distribution, etc.
In recent years, the fuzzy CFAR detectors have been widely studied, but they are mainly used for the multiple sensors.
Recently, a new CFAR detector based on fuzzy statistical normalization for non-Rayleigh distribution data, called fuzzy statistical normalization CFAR (FSN-CFAR) detector, is proposed by the researchers from the Institute of Acoustics of the Chinese Academy of Sciences. It proves to be a robust constant false-alarm rate detector for non-Rayleigh radar or sonar data both in nonhomogeneous and homogeneous environments.
The detector carries out the detection with two stages. At first, it does background level estimation based on fuzzy statistical normalization. In the following, it conducts signal detection, based on the original data and the defuzzification normalized data.
To evaluate the detection performance of the FSN-CFAR detector in K-distribution background, several computer simulations have been done. Multiple target situations and the case in which a clutter edge is present in the reference widow are considered.
Simulation results show that the performances of the smaller of CFAR and the order statistic CFAR detectors are poor in the case of a clutter edge. The greater of CFAR offers good performance in a clutter edge environment, but its detection performance is poor in multiple target situations.
The performance of cell-averaging CFAR (CA-CFAR) is the best in homogeneous background. However, the FSN-CFAR outperforms it in nonhomogeneous background and offers low constant false-alarm rate loss in homogeneous background.
The superiority of the proposed FSN-CFAR detector is perceptible, compared with other constant false-alarm rate detectors.
This research was supported, in part, by the National Natural Science Foundation of China under Grants No. 61222107 and No. 61431020.
References:
XU Yanwei, HOU Chaohuan, YAN Shefeng, LI Jun, HAO Chengpeng. Fuzzy Statistical Normalization CFAR Detector for Non-Rayleigh Data. IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS (Vol. 51, No. 1, pp. 383-396, January 2015). DOI. No. 10.1109/TAES.2014.130683
Contact:
XU Yanwei
Institute of Acoustics, Chinese Academy of Sciences, 100190 Beijing, China
Email: xyw@mail.ioa.ac.cn