The invention discloses a polarimetric synthetic aperture radar image segmentation method based on a two-stage region merging strategy, which aims at improving the segmentation accuracy of a non-uniform region, avoiding excessive complex operation and improving calculation efficiency. The technical scheme is that the PolSAR image is segmented into super pixels, and then Wishart merging and KummerUmerging are carried out in turn. The Wishart merging phase merges the original superpixels which clearly come from the same object quickly, and the KummerU merging phase merges the remaining regionsiteratively to obtain the final segmentation results. The method can improve the accuracy and computational efficiency of the merging, and combines the KummerU energy loss, the edge penalty function value and the homogeneity penalty factor to obtain a new measurement standard, which can further improve the segmentation accuracy.