Surface splicing defect analysis and application of polarization maintaining fiber using graph cut with illumination priors

Abstract

A new surface splicing defect analysis and application technique of Polarization Maintaining (PM) fiber is proposed. In contrast to the traditional artificial experience based analysis method, we not only develop an automatic defect segmentation technique for the fiber splicer but utilize the image features of splicing defect to assist evaluation of the splicing effect. First, we employ a standard fiber splicer to implement the splicing operation. Both the visible and the hot (infrared) images are captured during that processes. Second, we use the image processing techniques to analyze the image features for both the visible and the hot images. The Hough line detection is used to monitor the core-offset or the angle tilt problems of spliced fibers in visible image. A new Graph Cut Model (GCM), which uses the Multivariate Gaussian Mixture Model (MGMM) as the illumination prior of transmitted rays, is employed to segment the splicing defect in hot image. Third, multiple defect image features, such as the linear edge, the defect shape, and the inertia moment are all computed for the description of defective region. Finally, the SVM classifier is employed to evaluate the fiber splicing effect. The defect features, the splice loss, the extinction ratio, together with the final precision output of optical component are utilized to train the SVM. By using this method, a reliable quality control measurement for the optical component of aerospace optoelectronic apparatus is developed. Many experiments have verified the validity of the proposed method.