The Research On Surface Defect Detection Of Mobile Phone Battery Based On Machine Vision

In the current industrial manufacturing,the appearance quality of most products,especially electronic products,is manual inspection,which cannot meet the development requirements of intelligent and integrated industrial production lines.In the 21 st century,the theory and technology of "artificial intelligence" in the fields of language recognition,image recognition,machine vision and so on have been relatively mature,and the integrated system combining machine vision and quality inspection of automation products is gradually developing.The research object selected in this paper is the large amount of use of mobile phone batteries,and the machine vision-based algorithm is designed to perform non-destructive testing on the surface defects of mobile phone batteries.First of all,the background and significance of the research is explained,the development of machine vision technology and machine vision-based defect detection technology and system at home and abroad are elaborated,and the content of the overall surface defect detection overall program and software testing tools is explained.Secondly,according to the characteristics of large resolution and tilt of the picture,the tilt correction and area of interest(ROI)extraction operations are designed to obtain pictures with relative horizontal position and to distinguish the battery area and background,which facilitates the convenience of subsequent processing and the improvement of detection speed.Based on the characters on the front of mobile phone battery,a supervised character recognition method based on artificial neural network(ANN)is designed.A template matching method and a minimum circumscribed rectangle search method are proposed,each character position of the ROI is determined,a training sample is prepared,and the character is extracted and predicted using ANN.The pixels of the minimum circumscribed rectangle are then modified to ensure that the front and back sides of the battery maintain the same texture.In this paper,the ROI after character processing is divided into two parts,and the sub-image traversal of the image after the character operation is proposed by the adaptivethreshold brightness method.The defective sub-image of the overlapping area is merged and the area where the naked eye detection is not defective is filtered out.That is,the defect area of the battery surface is determined,and then pixel distribution regularity feature of binary images trained are extracted by using the multi-class classification method of the SVM algorithm,and the trained defect types of the battery appearance are identified.On the Windows platform,the VS2015 compiler,OpenCV visual library and MFC are used to design the visual interface of the mobile phone battery appearance defect detection system.The test results are verified in the software interface by using industrial images with good hardware from the factory.The best variable parameters are determined from the perspective of identifying the rate and accuracy of detection.Through a large number of image verifications,the program can effectively detect character types,defect areas,and identify battery defect types.