An automatic defect-inspection method for optical isolators using image analysis

Abstract

Abstract Manual inspection is still widely used for defect detection in optical isolator factories. The manual method is not only inefficient, but also low reliability, and easy to be disturbed by visual fatigue. This paper proposes an automatic defect detection algorithm for optical isolators, which is an extension to a previous conference paper, with more algorithms and detailed description of the detection algorithm added. In the focusing part of the algorithm, an improved Laplace operator which increases the attention of oblique gradient is adopted. In the detection part of the algorithm, the image-enhancement based on adaptive threshold, the line and ring detection based on Hough transform, and process the detection results by clustering method are used. Experiments show that the scheme proposed in this paper can inspect a series of defects on the crystal area of the optical isolator in real time, the manpower demand can be reduced to less than 10 % and the detection accuracy is 93 %.