Research On Visual Detection System Of Plane Glass Contour Size

The detection of the plane glass profile size is one of the important technical measures for measuring whether the plane glass is qualified.At this stage,the detection of the glass profile size mainly relies on manual detection or simple mechanical detection,but they have low detection efficiency,high the labor intensity,and low precision.To solve these problems,this article based on the advantages of machine vision technology such as high accuracy,high speed,and non-contact,a plan for visual detection of plane glass contours is proposed.Its main research contents and results are as follows:First of all,according to the basic requirements of plane glass size detection,a basic technical solution of visual detection system is proposed.Through the technical requirements and experimental verifications,the selection of hardware and the arrangement of the light source are completed,and the establishment of the detection system is completed in the laboratory environment,finally achieved a clear glass image acquisition.Secondly,for the purpose of how to extract accurate glass edges,through theoretical analysis and experimental comparison firstly,it is determined that binarized glass images are filtered by the improved median filter algorithm and the maximum variance between classes.Secondly,There are phenomena such as noise interference and edge discontinuity in the processing of glass images for commonly used edge detection methods.This article proposes a scheme based on morphological edge detection and extraction methods,and using polynomial difference method to obtain sub-pixel edges based on morphological edge detection and extraction,thus it improves detection accuracy.Then,this article designed a glass shape matching algorithm based on Hu invariant moments,achieved the identification of glass shape types,this method is proved to be reasonable and practical through specific experiment.In addition,according to how to achieve the problem of size detection,this paper designed a glass profile registration algorithm based on the centroid-contour point distance,which achieved the detection of the maximum size difference,and verifies the real-time,accuracy and reliability of the method through experimentsFinally,the camera calibration and detection system testing and error analysis are completed.Using the calibration plate to get the camera’s parameters and the relationship between the pixel size and the actual physical size.Using the established detection system platform,the glass image processing is completed,the glass outline size data is obtained,and the error factors that may affect the accuracy of the glass outline dimension detection are analyzed in detail.The plane glass profile size visual detection system studied in this article can meet the glass profile size detection of plane glass with a size no larger than 300mm X 200mm,and the detection accuracy is 0.2mm,it laid the foundation for the follow-up research and development of glass profile size detection automation equipment.