Surface defect detection method for glass substrate using improved Otsu segmentation

The image quality degradation caused by noise makes the automatic optical inspection of surface defects difficult. This paper develops a method based on thresholding segmentation to detect the surface defects in a glass substrate. Traditional Otsu segmentation has poor anti-noise ability. In order to improve the traditional Otsu method, a straight-line intercept histogram is established directly from the two-dimensional information of an image, and then Otsu criteria can be used to find the best intercept threshold from the one-dimensional histogram established. The improved Otsu algorithm not only is simpler than the two-dimensional Otsu methods, but also has a robust anti-noise ability. In the surface defect detection, the contrast feature between object and background is simply extracted after the segmentation based on the improved Otsu method, and surface defects can be decided by the threshold of the contrast feature. The data used in the experiments include the surface images acquired by a line-scan CCD camera. The experimental results demonstrate that the proposed method is effective and computationally efficient.