The internal defects of the workpiece is difficult to test. Since the image recognition technology can accurately determine the work-piece internal defects and define the internal defects of the work-piece areas, the tomography image segmentation technology has been widely used in the internal defect detection of work-pieces. In order to overcome over-segmentation phenomenon during adopting conventional watershed algorithm to segment the image, this paper proposed a sort of intelligent fusion algorithm of image edge extraction. It could firstly smooth the image with aid of opening-closing algorithm based fuzzy mathematical morphology. Based on the mathematical morphology, it could secondly compute the gradient operators, and finally, it segmented the gradient image to obtain the expected image. The comparative experiment study of simulation demonstrated that it is better in eliminating over segmentation phenomenon, better in real time and more applicable in image recognition. The research result showed that the proposed intelligent fusion algorithm has great reference significance to improve the image processing quality.