Sorting potato by shape is an important way to improve the economic value of patato. This paper presents an image edge information based detection method. Firstly, the original BMP image was grayed by using the formula of “R+G-B”, and then the gray image was changed into binary image after open and close morphological processing; later, the edge was detected by ‘canny’ module; finally, the maximum transverse diameter (L) and the maximum longitudinal diameter (LD) were defined with the edge information. And the value of L/L_D was selected to be the shape characterization parameters. The test results show that this detection method can be effectively applied into potato classification, specifically for online detection of potato shape.