Abstract: To investigate the segmentation and recognition of complex target whose gray has stochastic distribution, a new edge detection method, which used the gray levels of the pixel to compare with those of its eight neighbors, was presented to segment the target. After filling in the edge pixels and eliminating image disturbances, a fast polygon approximation algorithm was implemented to get the polygon of the target with fixed angles for recognition. The approach solved the difficulty of segmening such a complex image and got a good result of segmentation and recognition. It is effective and robust to detect the edge and recognize the target with the approach presented above.