On-line Detection Technology And Application Of Commutator Plane Defect Based On Machine Vision

With the continuous upgrading of the manufacturing industry,and constantly moving in the direction of automation and intelligent development,the application of machine vision is an important part of the manufacturing upgrade,the use of machine vision for automatic detection of products in the product manufacturing process by more and more Multi-enterprise favor.In this paper,the motor commutator of the automated manufacturing production line as the background,the study of commutator multiple surface defects visual inspection part of the main research are as follows:The existing defects are reclassified into surface defects,linear defects and color defects.According to the characteristics of the defect image,different pretreatment methods are designed.In the image of the two end faces of the commutator,the barking part adopts PM filter,and the other regions use the mean filter according to the actual situation.In view of the noise generated by the color camera in the detection of the straight hook,the median filter is used to complete the noise reduction.In order to detect the noise produced by the color camera in the detection of straight hook,the median filter is used to complete the image noise reduction.The detection image of the commutator needs to be divided into different ROIs: In the segmentation of the end face image ROI,the inner circle detection parameters are used as the reference,combined with the size parameters of the commutator,the image is divided into four areas: bakelite,stiffeners,copper shells and hooks.At the same time,in order to facilitate the defect measurement also need to carry on the polar coordinate transformation to each region.Aiming at the darkness of the bakelite area,an enhancement method for the bakelite ROI image is used by cutting the middle section of the accumulated histogram.For the ROI of the positioning groove,the hook top and the straight hook side are directly split with the method of template matching,and finally the statistical matching data are used to verify the rationality of the algorithm threshold selection,and the reliability of the algorithm is verified by three aspects: precision,speed and matching rate.A method of automatic threshold segmentation and local dynamic threshold segmentation based on histogram is proposed for the detection of planar defects and linear defects in grayscale images of various commutators.And then through the Blob analysis,measuring the characteristics of the various regions,to complete the quantitative and qualitative of the defect.In order to solve the problem of inaccurate problem of linear feature threshold segmentation,a method is proposed to extract the ridge line of the linear region,in this way,the analysis and measurement of the defects are completed.Finally,the recognition results of the two types of defects are identified,and the recognition rate is up to 97%,verifying the effectiveness of the algorithm.For the color defects,which can not be identified in a gray-scale image,exist in the lower end of the copper shell area and straight hook area.The defects,which color of tin is abnormal,is showed in the conditions of low ring light.With the use of support vector machine,we have completed the two categories that have the defect and no such defects.Under the coaxial light condition,the SVM based on the fusion feature is trained to complete the classfying of the three kind of samples,which are no defect,imprinting and black.The statistical recognition rate is 98%.