Research On Key Technology Of Tool Gap Detection System Based On Machine Vision

The electrode burr caused by micron gap of cutting tool edge is a serious potential safety hazard source of power battery.At present,due to the lack of high-precision automatic detection equipment for cutting tool gap of lithium battery electrode,the quality inspection personnel mainly observe the microscopic image to determine the tool wear subjectively.This method has the problems of low detection accuracy,poor repeatability,high detection labor cost and large visual impairment,which can not meet the requirements of enterprises for detection accuracy and speed.Therefore,in this paper,the key technologies of highprecision detection system for tool gap are studied.(1)An attachment removal network ARNet for micro image of cutting edge is proposed.In view of the adhesion of the tool edge,the detection accuracy is reduced,and the existing attachment removal algorithm has the problem of incomplete tool edge recovery,or even removal failure.Firstly,the binary mask guidance module is used to distinguish the target and background features,and then the removal process learning module is used to extract the timing information in the recursive process;then the self attention precise separation module is used to establish the dependency information of the attachment features in the global,so as to accurately remove the attachment.The experimental results show that compared with prenet,the PSNR and IOU are improved by 1.016 d B and 3.48% respectively,and the real edge of the tool is restored more completely.At the same time,86.5% of the parameters and 90.9% of the computation are reduced.(2)A sub-pixel edge detection algorithm of tool image based on improved Zernike moment is proposed.In order to solve the problem of virtual tool edge caused by light source diffraction and low accuracy of traditional edge detection algorithm,a larger 7x7 pixel template is used firstly;secondly,according to the gray distribution of the transition interval between tool edge and light source,a three-level gray edge step model is used;then,Otsu algorithm is used to optimize the sub-pixel edge determination conditions.The experimental results show that this method has higher detection accuracy than the traditional edge detection algorithm,and realizes the accurate positioning of the tool edge.(3)A tool gap detection algorithm based on improved region growing is proposed.In view of the problem that the adjacent gapes are conglutinated,which leads to the virtual height of gap parameters and seriously affects the accuracy of the measurement results of the tool gap,firstly,based on the difference results of the tool edge curve,an adaptive Valley seed selection method is proposed;then,the domain search model is improved to search,and the pre selection is generated by combining with the difference sensitive growth conditions Finally,the connected preselected gap regions are merged according to the topological structure to optimize and output the gap location interval.The experimental results show that the proposed method can detect a single defect according to the gap topology,the gap starting position is accurately divided,and the average detection time of4096x2168 resolution image is 0.0232 s,which meets the real-time requirements.