Abstract: In this paper, a background statistical characteristics based robust sonar target constant false alarm ratio (BSCR-CFAR) detection algorithm is proposed to deal with the target detection problem in complex non-homogeneous underwater environment. In the proposed BSCR-CFAR detector, the automatic censored mean level detection (ACMLD) and order statistic CFAR (OS-CFAR) detection algorithms are applied to the variability index CFAR (VI-CFAR) detection algorithm, and then the more matched CFAR detection algorithms are adaptively selected by the assessment of background statistical characteristics. The simulation and the analysis results of sonar measured data indicate that by comparing with other detection algorithms, such as cell average CFAR (CA-CFAR), greatest of CFAR (GO-CFAR), smallest of CFAR (SO-CFAR), OS-CFAR and VI-CFAR, the performance of the proposed BSCR-CFAR method for CFAR detection still maintains better robustness in typical non-homogeneous environments, such as reverberation edge, reverberation region, and one or more strong discrete outliers.