Research On Visual Inspection Technology For Display Defects Of New Array Panel

With the emergence of a large number of new display technologies,the way people interact with electronic devices has been greatly enriched.As the most important medium of human-computer interaction,the performance and quality of display screen directly affect user experience,so people have a higher demand for its display effects.However,the display screen inevitably produces some defects that affect its display effect due to the complicated manufacturing process and transportation process,and these defects will determine the quality of the screen.Therefore,it is very necessary to detect defects during the screen production process and before leaving the factory.Currently,screen defect detection is done manually,but manual defect detection has various limitations,so it is difficult to meet the needs of large-scale production lines.Therefore,relying on automated testing equipment to replace manual screen defect detection has attracted more and more research institutions,experts and scholars.In this paper,machine vision detection technology is used to further study screen defect detection.According to the project's expectations and design requirements,a hardware image acquisition system for screen defect detection was built to collect high-resolution screen images that are not disturbed by the external environment.This article investigates the types and manufacturing processes of display screens on the market,and analyzes the causes of screen defects.In view of the characteristics of screen defects,this article chooses suitable industrial cameras and lenses,external light sources,module generators and other hardware equipment.In order to improve the image quality,the image acquisition hardware system needs to be specially designed and arranged to avoid the interference of the external environment,and because the industrial camera will produce t he moiréphenomenon when capturing the screen image,it is also necessary to research and design the hardware device to suppress the moiré pattern.The algorithm detection in this paper is divided into three parts: image preprocessing,defect segmentation and defect classification.Screen image preprocessing is the basis of defect detection and will directly affect the result of defect segmentation.Preprocessing first eliminates the interference of dust and noise,then corrects and extracts the image detection area,and finally suppresses the background texture of the image.Defect segmentation is the focus of this paper.Designing a defect segmentation algorithm with high accuracy and good robustness is the research goal.Defect segmentation adopts the idea of background reconstruction,and two background reconstruction algorithms are designed from the perspective of image processing and deep learning: median background reconstruction algorithm and background recovery algorithm based on Pix2 pix model.Through the gray difference of the defect area between the detection image and the reconstructed background image,the point,line and mura defect segmentation of the screen body is completed,and the performance of the two detection algorithms proposed in this paper is analyzed and compared through experiments.Defect classification is the statistical analysis of screen defects to determine the subsequent processing of the screen and the feedback of the production process.This paper designs a lightweight neural network structure that satisfies online defect classification,and accurately classifies the detected points,lines,and mura defects.Finally,the existing classic classification network is verified and compared with the design model in this paper.The research results of this subject provide a feasible solution for the visual defect detection of the screen to overcome the limitations of the existing manual detection.It has certain guiding significance for the defect categories and detection methods of the screen,as well as the improvement of the subsequent processing and production process of the screen.