Research And System Implementation Of Fiber Optic Connector Defect Detection Based On Machine Vision

As an important part of fiber-optic communication system,fiber-optic connector will directly affect the performance of communication systems.The traditional fiber-optic connector detection method is manual visual inspection,which has low detection efficiency and poor real-time performance.In contrast,the machine vision inspection system has the advantages of high speed,high precision and zero damage,so this article aims to replace the traditional manual detection with machine vision technology to implement a fiber-optic connector defect detection system.The test object of this article is a 32-pin fiber-optic connector.Combined with the requirements of the fiber-optic connector defect detection system,the entire defect detection system is designed,and the hardware is analyzed and selected.In this article,LED light source for forward illumination of fiber-optic connectors,and image acquisition using CMOS industrial cameras.In order to avoid the interference of contour blurring on subsequent defect detection,this article enhances the image in the image preprocessing process,analyzes three commonly used image enhancement methods,and introduces evaluation indicators to evaluate various image enhancement methods,then selectes a better Laplace sharpening operation,but sharpening brings extra noise to the image.This article introduces the commonly used filtering method,and selects the filtering method to denoise the image according to the evaluation index,then uses the template matching method to identify the type of connector.Before the defect detection,it is necessary to extract the detection area,that is,the extraction of the jack.In this article,the local adaptive threshold segmentation combined with the morphological processing method is used to enhance the contour of the jack,and the Hough transform is used to locate the jack,but the positioning jacks are out of order and contain two positioning holes.Then the benchmark data comparison method is used to sort the jack information,and the jacks can be accurately segmented according to the sorted jack information,so as to facilitate subsequent defect detection.In order to perform defect detection on the jack map,this article firstly combines the characteristics of the coating layer to make a mask,and uses the fixed threshold segmentation and the geometric knowledge of the circle to obtain the center of the coating layer,and then calculate the eccentricity of the jack.The coating layer gray histogram of 50 fiber-optic samples is accounted.After analyzing the gray histogram,this article proposes the threshold segmentation method based on the 3? criteria for normal distribution to detect the defects in the coating layer,then saves the detected defect information for later interface display.Finally,the user-interface is designed and implemented,and the whole detection system is implemented.Then the accuracy and consistency of the system are tested.The test results show that the defect detection system implemented in this article can accurately detect the degree of eccentricity and contamination of the coating layer of each jack in the fiber connector,achieving the intended purpose.