Abstract:

A new automatic censoring based cell averaging constant false alarm rate (ACCA-CFAR) algorithm is proposed for target detection from high resolution synthetic aperture radar (SAR) images. First, the G0 distribution is used for statistical modeling of clutter to deploy its advantage of strong compatibility. Second, a global threshold prescreening method based on G0 distribution is presented to obtain index matrix for screening out interfering targets pixels in the sliding window(S-W). Through automatic censoring, the proposed method is therefore well adaptive to complex situations. Third, a fast algorithm based on integral image is developed. With the proposed algorithm, the computational burden is greatly reduced. Meanwhile, the calculation complexity is independent of the S-W size. Accurate target detection results are obtained via count filter and morphological processing. The proposed approach has two notable advantages, i.e., being fully adaptive while with much smaller computational cost compared with existent similar algorithms. Experimental results for TerraSAR-X images demonstrate its effectiveness and usefulness in practical engineering.