Abstract:

Considering the big error and low accuracy in traditional CFAR segmentation based on Weibull model, the paper proposes the improving methods that the samples of the specific azimuth are processed using vertical median filter before segmentation and the zone growth filtering method is used to filter the noise points. The improving methods are compared with the traditional CFAR segmentation and the counting filtering by using the segmentation evaluation standards of gray-level contrast and ultimate measurement accuracy. The experimental result using the MSTRA samples demonstrates that the new methods improve the segmentation accuracy and the segmentation results are superior to the traditional CFAR segmentation methods.

Key words: segmentation, vertical median filtering, zone growth, CFAR