Research On Optical Fiber Module Detection Algorithm Based On Stereo Vision

With the continuous development of the optical fiber communications industry,the demand for optical modules has increased dramatically.Especially,the market demand for Small Pluggable Fiber Modules(SFP)has increased sharply.Up to now,the defects detection methods of SFP in domestic manufacturers are still using traditional human eye detection methods.It's urgent to develop the fiber defect visual inspection system since the low efficiency,low accuracy,high labor intensity,and unstable detection quality of manual detection.Optical Fiber deformation detection is the core of the fiber defect detection system.The article is to detect the optical fiber deformation defect by using the binocular vision.The workload of the article could be divided to following parts:(1)Research the camera calibration algorithms.The article studies and compares several classical camera calibration methods As the first step in stereo vision.Meanwhile,the article has a detailed theoretical analysis to the mutual transformation and distortion processing of the coordinate systems in the camera's imaging model content.Especially for the Zhang's calibration method,the article analyses each step of the algorithm flow,and conducted a lot of experiments to verify the performance of the algorithm.(2)Implemented and improved the layered fuzzy gradient enhanced SURF feature extraction algorithm.In order to solve the problem of unsteady gradients in flat areas and instability in edge areas.Based on the advantages and disadvantages of feature extraction algorithms such as Harris,SIFT,and SURF,a hierarchical fuzzy function is introduced to perform gradient transform and Layered enhancement on SURF features.The improved algorithm can well remove the unstable points of the fiber module in the uneven and edge regions,and improve the accuracy of feature point extraction.(3)Based on BBF(Best Bin First)to improve the KD-Tree matching point search strategy.The article analyzed and compared a variety of traditional feature matching search strategies,and finally selected the KD-Tree search strategy to eliminate false matching points.At the same time,the article also proposed a stereo matching search strategy algorithm based on BBF improved KD-Tree for the limitation of KD-Tree dimension.Through experiments,the improved algorithm can effectively reduce the false matching rate.(4)Built an experimental environment and completed fiber module defect detection experimentsDuring the experiments,an experimental environment was established and fiber optic module defect detection experiments was completed.The point cloud information is obtained by point cloud calculation on the obtained partial image of the large scene.The depth,normal vector,and a series of parameters are accurately registered with the template map in the standard template library.So that,it could obtains a defect determination result.