Research On Defect Detection System Of Brake Oil Tube Head Based On Machine Vision

This paper mainly studies the inspection system of the upsetting defect of the brake oil tube head based on machine vision,which aims to complete the rapid and accurate judgment of all kinds of defects in the upsetting head of the brake oil tube combined with the advantages of the machine sense detection.According to the physical characteristics of the upsetting head of the brake oil tube,the construction of the image acquisition system is completed,including the light source,the lighting mode,the industrial camera and the lens selection,and the relationship between the four traditional coordinate systems and the lens distortion model are studied.Some defects in the practical application of the traditional Zhang's calibration method are analyzed,and the base is put forward.The self calibration algorithm of vanishing point and the verification of final data prove that the self calibration algorithm based on vanishing point is more suitable for industrial environment.The multi histogram sliding window algorithm is designed to complete the segmentation of the background area of the captured image,and an adaptive illumination compensation algorithm based on the field reconstruction is proposed to eliminate the uneven illumination caused by the factors such as outdoor illumination.In order to eliminate the high frequency white noise produced in the transmission process,the base is designed that an adaptive filtering algorithm for decision making.According to the difference of the area and characteristics of the defect distribution of the brake oil tube head,the defects of the upsetting head of the brake oil tube are divided into three categories.In view of the defect on the surface of the head of the brake oil tube,a paint detection scheme based on the HSV color space is designed.For internal contour defects,after the segmentation of the inner contour,a roundness discrimination and detection circle approach detection scheme is designed,and the external contour is extracted after the edge extraction.The outer contour is fitted by circular fitting,the absolute distance information of the two contour edges is calculated,and the thresholds of distance determination are set.At the end of discussion,the upsetting head detection system of brake oil tube is built with Qt,and the system testing system is completed.The results show that the real-time and accuracy of the detection system meet the actual detection requirements and achieve the desired purpose.