Research On The On-Line Detection Technology Of Connector's Screw Based On Machine Vision

Defect detection of products is one of ordinary jobs on flow-line. At present,the detection of products is done manually,which not only increases the works labor intension,but also is difficult to meet the demand for quality in modern industry. In the recent years,with the application of machine vision to industry detection,it has been a trend to apply computer technology to detect product quality.Detection technology based on machine vision were presented for defect connector's nut with the application of Matlab image processing tools after studying related scientific literatures. Threshold Segmentation algorithm based on effective gray pixel's statistical histogram and the iteration Hough transform algorithm of circle detection based on limited region (IHTLR) were presented after further study on related technologies. The main work includes:1. About the design for illumination system. Illumination from back and side light was adopted through comparison of merit and demerit of various illumination method, which not only extruded outline of products but also concealed and weaken the little puncture leaving in the frame part of products,which is of advantage to the following processing .2. The image pre-processing. During image pre-processing, we found that edge was better kept and noise were ideally removed if using median filter, through analyzing and comparing of effectiveness of various slide filters such as average filter,median filter and Gaussian filter.3. As for image segmentation and edge extraction. Threshold Segmentation algorithm based on effective gray pixel's statistical histogram was presented through comparison of common segmentation algorithms,and then edge was extracted with application of mathematic morphology.4. As for pre-detection of defect products. After analysis of characteristics of defect products, at first a part of products were pre-detected with the application of pre-detection method,and then the remaining products were accurately detected, which could significantly improve the speed of system.5. As for circle detection using Hough transform. After study of the traditional Hough transform and the randomized Hough transform (RHT),the iteration Hough transform algorithm of circle detection based on limited region (IHTLR) was presented, which was characterized by quicker speed of iteration,litter stock room and better precision of detection.