Research On Defect Detection System Of Transparent Materials Based On Computer Vision

In recent years,with the development of modern manufacturing industry and optical technology and the improvement of material technology,transparent materials have good application prospects and research value in the preparation process of precision optical instruments,lighting equipment and glass.Therefore,transparent materials should be improved in the processing quality,and artificial detection is difficult to meet the needs of modern production.This article is about In this background,the defect detection of transparent materials is studied and realized.Firstly,the research background and significance of transparent material defect detection are introduced,and the overall design scheme of transparent material defect detection system based on computer vision is proposed.The system is divided into software module and hardware module.The hardware module is divided into three parts.After confirming the hardware selection,a set of transparent polarization image acquisition device is designed,which provides the data base for the following image processing.At the same time,the overall flow framework of the software module is proposed,and the function of the transparent material defect detection system is clarified.This paper studies from four modules.First of all,we build a transparent material defect detection device based on computer vision to obtain the polarization image of transparent material.The second is to preprocess the image,including the specific operations such as gray level change,gradient reconstruction,image adding light,image smoothing,etc.after the above preprocessing,the defects in the polarization image are largely separated from the background.Then,the data of the obtained polarization image is enhanced and labeled in this paper,through the way of image rotation,translation,zooming,changing the light and shade,etc The original image data is expanded to form the polarization image data set of transparent materials.Next,the paper studies five common defects of transparent materials,using the target detection model Yolo v3,and tries to combine the intersection and parallel ratio and the improved non maximum suppression algorithm optimization model on the basic network structure.The detection accuracy of defect location on the improved Yolo v3 is 98.7%,and the detection accuracy of defect type is 90.0%,compared with Yolo v3 improved 4.6% in defect location detection and 3.8% in defect category detection.The software processing module of this paper is implemented on the platform of Ubuntu 16.04 by using python programming language and Open CV function library.It has high efficiency of defect detection algorithm and can effectively detect defects.It has a certain significance to promote the defect detection of transparent materials.