Abstract: False welding of data lines is a common defect in the welding process of data lines, which seriously affects the acceptance and subsequent use of data lines. At present, visual inspection is often used to judge such defects, which results in low work efficiency and high missed detection rate. Existing algorithms were mostly used to detect defects such as welds and holes on metal surfaces. However, there are few specialized methods for detecting the welding quality of data lines. According to this, a method of data line welding quality detection based on machine vision was designed. First, the original image was segmented to obtain the image of the area to be detected by corner detection, and then the area to be detected was segmented by color image local binary patterns (LBP). Then, contour detection and morphological operations were used to obtain the contours of each region, and welding defects were classified based on contour features. Finally, support vector machine (SVM) was used for classification and statistics. The experimental results showed that the proposed detection method had a defect classification accuracy of 96%, strong pertinence, simple operation, and high practicality.