Research On Defect Detection Method Of Mobile Phone Screen Based On Computer Vision

In recent years,technological progress and the popularization of the Internet have led to the continuous increase in the number of mobile phone users,and the production demand for mobile phone screens is huge.Mobile phone screen defect detection is an important part of production,and its quality control and quality inspection issues have attracted much attention.Traditional quality inspection methods mainly rely on quality inspectors to identify production defects with the naked eyes,which is not only inefficient,but also difficult to identify with the naked eyes for some small or micro defects.The rise of machine vision provides new ideas for defect detection of mobile phone screens.People apply machine vision-related algorithms to defect detection of mobile phone screens,which liberates manpower to a certain extent and improves detection efficiency.However,the application scenarios of machine vision are limited and can only be used in low-noise,well-lit scenarios,and the algorithm has poor stability and large adjustment parameters.With the rapid development of deep learning theory,it is found that deep learning performs better and better in defect detection tasks,with fast detection speed,high accuracy,and wider application scenarios.Therefore,this thesis studies the detection method of mobile phone screen defects combined with deep learning,and realizes an algorithm that can accurately identify mobile phone screen defects and defect types.The main contents of this thesis are as follows:(1)This thesis proposes an improved fusion defect detection algorithm suitable for mobile phone screen defects.Firstly,various image preprocessing operations are performed on the defective image of the mobile phone screen to solve the problems such as noise and blur in the image itself,so as to realize the preliminary determination of whether the mobile phone screen has defects.(2)In view of the problem of small samples in this thesis,this thesis proposes an improved image segmentation algorithm based on the idea of sliding window;proposes a method for generating virtual samples of mobile phone screen defects based on the ACGAN algorithm;proposes a method based on data enhancement to generate The new defect sample method extends the mobile phone screen defect dataset.(3)Based on the improved Faster RCNN and YOLO v5 algorithms,this paper proposes a deep learning method suitable for mobile phone screen defect detection.This paper improves the NMS method and ROI Pooling method,simplifies the backbone network,and introduces an attention mechanism,so as to achieve a better recognition effect for mobile phone screen defect detection.This thesis verifies the effect of the proposed computer vision-based mobile phone screen defect detection method through experiments,and realizes the preliminary judgment of mobile phone screen defects.The accuracy rate of mobile phone screen defect type detection has reached more than 98%,and the detection time of a single defect is 0.012 s.