Algorithm for Locating and Detecting Parallel-edged Sewing Thread of Irregular Small Cloth

The quality inspection of sewing thread plays an important role in the intelligent garment industry. In the manufacturing process, part of the cloth needs to be processed into small size parts on the ready-to-wear garments, so the sewing thread must be close to the edge of the cloth during the stitching process. This type of sewing thread is commonly referred to as parallel-edged sewing thread. Because the parallel-edged sewing thread is too close to the edge of the cloth and the trend of the stitch is irregular with the change of the shape of the cloth, it is difficult to locate the position of the sewing thread and detect its quality. In order to solve this problem, an algorithm to locate and detect the parallel-edged sewing thread of irregular small cloth is presented. First, the image is filtered through a multi-angle and multi-scale Gabor filter to extract linear texture features. Then the color extraction process based on the HSV color space is introduced to obtain the mask area of the small cloth in the image and the contour discovery algorithm is introduced to extract the peripheral contour point sequence. The polygon fitting and polygon shrinking operations are performed on this sequence to obtain a mask area, and then the intersection and merge operation is performed on this mask to complete the positioning of the sewing thread area. In this area, a mask-based improved moving window local binarization algorithm is introduced to process the image and the connected domain algorithm is introduced to complete the quality detection of the sewing thread. After experimenting with 800 pieces of even stitching data sets, it is found that the algorithm can accurately locate the area of the parallel-edged sewing thread and detect the quality of the sewing thread, and the calculation speed is 0.2 to 0.4 seconds per sheet, which significantly improves the efficiency compared with manual inspection.