The geometric active contour（GAC） model can avoids reparameterization based on the geometry characteristics， but it has the poorly ability in dividing a fuzzy boundary. In contrast， by maximizing the grayscale difference between the target and the background， the Chan-Vese（CV） model can effectively differentiate the fuzzy boundary. Based on the aforementioned consideration， the GAC-CV hybrid model is proposed， where the edge of the image and the region information are merged into the same "energy" function， and the various segmentation strategies are adopted for different segmentation targets to improve the capture ability of the concave edge.The segmentation results about thewater-repellent images of seven grade of the insulators show that the hybrid model has superior segmentation performance， and the detection rate of water bead highlights is as high as 95%.