Abstract:The pit on the cylindrical surface is one of the important indicators of cylindrical coated lithium battery circumference surface defect detection. There are two interference factors in the detection of craters on the circumferential surface. The image has uneven brightness along the circumferential and axial directions, and oil contamination on the circumferential surface. To solve these problems, a method for pit defect detection based on machine vision is proposed. Firstly, the gray scale distribution curve is extracted along the axis of the circumference. Then, the gray level difference model is used to extract gray discontinuous points in gray distribution curve, which is not sensitive to light distribution and oil contamination. The extraction threshold of discontinuous points are determined according to the reflective feature of the circumferential surface. In this way, the pit defect detection is achieved. The algorithm is evaluated on the selfbuilt image database SUTBY. Experimental results show that false rejection rate and false accept rate are both 0 percent. The actual test results indicate the uneven brightness has no influence on pit extraction, and there is no false detection caused by oil pollution.