Research Of On-line Defect Detection Of Button Hole And Color Difference Based On Machine Vision

The button is an important part of clothing. In the manufacturing process of button, there are many defects such as hole's deformation, cracks and color difference. At present, the defect is detected by human vision, many enterprises are eager to use machine vision to replace human vision to save the cost of human resources. According to the above requirements, this thesis proposed a button hole and color defect online inspection algorithm, using the smart camera system with TMS320DM6437 DSP core as hardware platform.For button hole's deformation defect, use the adaptive threshold algorithm to segment the hole area. By calculating each hole's area size and roundness, we can identify which hole is qualified and which is not. For those crack buttons, use Canny edge detection algorithm to detect the edge of crack. Taking advantage of the crack's characteristics of perpendicular to hole's contour, we calculate the contour's area size and its minimum bounding rectangle's ratio of width to length, then we can identify whether the hole is qualified. The color difference defect detection algorithm is an on-line detection combined with an off-line learning process. In offline learning process, we use K-Means clustering on the template image to get the color information of the button. In online process, we use the location of button holes to match the template image, and detect each zone's color difference according to the template image.We established a test databased to assess the algorithm on defects inspection of the hole deformation and cracks. The holes deformation and cracks on average algorithm running time is 92 ms, the false negative detection rate was 3.72% and the false detection rate is 3.66%. We test the chromatic aberration detection algorithm on a small sample, the false negative detection rate was 7.4% and the false detection rate is 13.95%.