Research On Glass Surface Defects Inspection Technology Based On Machine Vision

At present,there are few studies on the glass defects in the process of flat glass re-processing,and there is no mature automatic detection systems applied in practice.The detection system based on machine vision not only has the advantages of non-contact and automatic,but also has better reliability,faster speed and higher quality than manual inspection.In this paper,to achieve the goal of glass surface inspecting automatically,a series of in-depth study with related theory and implementation method are carried out.The main research works are as follows:First of all,by analyzing the requirements of the visual inspection system and the detection difficulties faced in flat glass re-processing,a new overall scheme of the glass surface defect visual inspection system was developed.The hardware structure of the detection system and the selection of the key hardware(light source,camera)were completed.Secondly,through the literature analysis and experimental verification,three kinds of glass defect acquisition systems with good collection effect were proposed: glass surface defect acquisition system based on grating,glass surface defect acquisition system based on intelligent light source,and low surface illumination glass surface defect acquisition system.And by comparing and analyzing the acquisition results of the three systems and their advantages and disadvantages,a theoretical basis and experimental verification for the selection of image acquisition system were provided.Finally,the image processing algorithms were studied by means of experimental contrast analysis for the acquisited glass surface defect images.Mainly includes the research on the methods of removing noise,uneven illumination and background fringes in glass defect images.And two effective threshold segmentation methods for inhomogeneous glass defect image were studied,threshold segmentation method based on edge improvement and threshold segmentation method based on gray-scale fluctuation.The extraction and feature analysis of connected regions were studied,and a new method based on shape feature was proposed.The experimental results show that the threshold segmentation method based on gray-level fluctuation transformation can greatly improve the segmentation effect of glass surface defects,and the glass defect classification method based on shape feature has high classification accuracy,to a certain extent,both of them meet the actual inspection demands.This paper has completed the overall design of the system,the research of image acquisition system and image processing algorithms,which provides a theoretical basis for the further improvement of the vision detection system.