Automobile Glass Size Detection Based On Machine Vision

As one of the important parts of automobiles,automobile glass has strict size standards in the production process,and its size inspection is an indispensable part in the production process.Automobile glass has a variety of shapes and sizes.In the production process,manual detection methods are mainly used.The detection efficiency of this detection method is not high,which is not conducive to the automated production of the factory.The detection technology based on machine vision is a non-contact measurement technology with fast detection speed and high accuracy,which can realize online detection and real-time analysis and judgment.Therefore,the development of a system for on-line inspection of automobile glass has very important practical significance for improving the production efficiency of factories.Aiming at this problem,this paper researches and designs a set of automobile glass online inspection system based on machine vision.The research contents of this article are as follows:(1)The system technical requirements are analyzed,according to the design requirements,completes the system hardware design and selection,the hardware structure of the online inspection system for automobile glass size is completed.(2)Aiming at the problem of ambient light interference during the glass image size detection process.Multiple filtering algorithms are tested experimentally by using three evaluation indexes of MSE,PSNR,and SSIM.It is found that median-based and Sigma filtering can maintain edge information while denoising,and used gray-scale linear stretch enhancement to lay the foundation for subsequent high-precision detection.(3)In order to improve the accuracy of size detection,a glass edge sub-pixel detection method based on the gradient edge model Zernike moment is proposed,and use the bilinear interpolation to select the model edge width parameters,the sub-pixel edge contour of automobile glass is obtained.Finally,combining the early termination strategy and the image pyramid search strategy,the glass profile to be detected and the template glass profile are registered and calculated to increase the registration speed.(4)Designed and developed automobile glass online inspection software.Comparing the detection system with the industrial field manual multi-point statistical process(SPC)statistical process data,the average detection error is less than 0.04 mm,and the repeated variance is less than 0.03.The experimental results show that the system developed in this paper can accurately and reliably detect the size of automobile glass.,The detection accuracy of the system can meet the actual detection needs of the factory.