Research On Visual Inspection Method For Surface Defects Of Large Size Cabinets

Surface quality is an intuitive characterization indicator of cabinet quality.At present,manufacturers pay more and more attention to the cabinet surface quality because customers require higher quality of electrical products and the products meet the increasing market competition.Accordingly,surface quality inspection technology of cabinets has been rapidly developed.In particular,machine vision detection technology,which is intuitive,fast and low cost,has been widely used in the surface quality inspection of industrial products.Moreover,vision-based detection technology is helpful to improve the automation level in industrial applications,making it become a research hotspot of cabinet surface defect detection.However,most existing methods of surface defect detection on electronic components and steel plates are difficult to address the problems of uneven illumination,complicate surface topography and edge defect inspection,making it cannot be directly applied to detect the surface defects of large-size cabinets.Therefore,this paper presents the key technology based on machine vision for surface defect detection of large-size cabinets.The research can provide a new method for appearance quality inspection of large-size products,which has important value in engineering applications.The main research contents are listed below.(1)The types,causes and characteristics of common cabinet surface defects are analyzed.In order to acquire high-quality images of cabinets,optical imaging theory is need to first study.According to the technical requirements of defect detection,a binocular vision image acquisition system is built,which meets the detection requirement on the cabinet surface in the size of 1.8m × 0.8m.The image acquisition frame rate is 5 frames per second(FPS)with the resolution of 3840 × 2748 pixels,which can detect the minimum defect of 0.5mm × 0.5mm.In addition,the cabinet defect detection system software is designed and written independently to realize automatic image storage and intelligent defect detection.(2)Feature-based image registration is investigated to detect the defects on special areas of the cabinets.The homography matrix among images is solved by image feature point extraction and matching.The spatial transformation relationship between the test image and the template image is obtained.By doing so,the image registration of the test image and the template image is implemented.Afterwards,the special regions specified by the person in advance are segmented.Finally,the weighted maximum inter-class variance threshold segmentation method is integrated with the image difference method to detect the defects on the segmented special areas in the cabinet surfaces.(3)Defect detection on background area using gradient threshold segmentation is proposed.This method can restrain the influence of uneven illumination to improve detection accuracy on large-size cabinet surfaces.Firstly,cabinet images are enhanced using the hyperbolic tangent curve transformation on image background.Then the Sobel operator is used to obtain the gradient image of background areas.Finally,the fixed threshold segmentation is employed to separate the defects on background area of the large cabinets.Experimental result shows that the detection accuracy(Precision)reaches 0.90,the recall rate(Recall)is 0.92,and the Fmeasure of the comprehensive accuracy rate and the recall rate is 0.98.(4)Edge defect detection on cabinets is researched by image block abnormal feature recognition.To this end,edge defect detection is implemented by utilizing the characteristics of the abnormal gray distribution of defects.First,the edge areas of cabinets are extracted and divided into multiple consecutive image blocks.Then image moment is constructed to acquire the grayscale long-axis feature of the edge image block,which is used as the discriminant indicator of abnormal image blocks.Finally,the Gaussian model-based anomaly is applied to build the recognition model of normal and abnormal image blocks to realize intelligent defect detection on the edge areas of cabinets.Experimental result shows that the accuracy of edge defect detection is 1.00,the recall rate is 0.79,and the F value is 0.88.In this paper,the visual inspection theory and key teniques of surface defects of large-size cabinets are investigated.Image acquisition system of cabinet surfaces is built.The design scheme and image processing algorithm of large-scale cabinet surface defect detection system are proposed.Meanwhile,image acquisition and image processing algorithms are designed and developed to establish an integrated software system,which meets the technical requirements of surface defect detection put forwarded by enterprises.