Detection And Classification Of Tiny Parts Based On Machine Vision

Machine vision technology has the advantages of stability and speed in detection,and is widely used in the manufacturing industry.This thesis mainly studies the machine vision-based copper foil buckle detection and classification system,which aims to use the inspection advantages of machine vision to complete the classification of copper foil buckle defective products and qualified products efficiently and accurately.The main content of this paper includes four aspects: building image acquisition module,object positioning method,detection and classification method and actual test.In terms of image acquisition,an image acquisition module is built according to the surface characteristics,detection accuracy and installation position of copper foil buckle.The hardware part includes the selection of light source,industrial camera and optical lens.Hikvision's MV-CE050-30GM5 million CMOS industrial camera,MVL-MY-1-65C-MP1 times telecentric The lens and white coaxial light source was choosed through demand analysis and experimental debugging,maximize the imaging quality.In terms of software,HALCON and Winform window application based on C# language are used to complete the image acquisition of copper foil buckle after configuring the network environment.In the aspect of object location,the illumination compensation algorithm based on low-pass filtering is used to solve the uneven illumination caused by external factors,and the image is reduced by bilinear interpolation to reduce the resolution of the image to reduce the amount of image data calculation.Two template matching positioning methods based on NCC and shape are designed to complete the precise positioning of the copper foil buckle.A comparative experiment was conducted,and the results showed that the degree of surface defects on the copper foil buckle would affect the accuracy and duration of matching positioning based on the NCC template,reducing the efficiency and accuracy of the classification system.Therefore,the shape-based template matching positioning method is more suitable for the detection and classification system.In terms of detection and classification methods,a defect detection based classification method was designed according to the quality requirements of copper foil buckle,and twodefect detection methods based on MLP classifier and difference model were compared in this method.in the classifier model,filter the image Laws texture,use the gray level co-occurrence matrix to obtain the Energy,Correlation,Homogeneity,Contrast feature statistics,and use the sobel operator Boundary and calculate the absolute frequency of gray value.Comparing the two results,the MLP classifier is superior to the difference model in accuracy and validity of program execution.In terms of practical testing,combined with Winform program and HALCON algorithm library,a user interaction module and a defect detection module were designed and implemented,and a copper foil snap detection classification system was integrated with the image acquisition module.The test results show that the false detection rate is 0.43% and the false detection rate is 0.The average detection time of the system for each copper foil buckle is 536.75 ms,which is lower than the factory required 600.00 ms,so the accuracy and real-time of the classification system Both meet the actual requirements.The detection and classification system of copper foil buckle completed by machine vision technology can complete the automatic sorting of copper foil buckle,which can meet the requirements of enterprises.The system can be popularized and applied in the defect detection classification of other small parts with similar characteristics.