Research On Size Measurement And Defect Detection Of Watchband Based On Machine Vision

In recent years,the sales volume of watchbands has become higher and higher.As replaceable watch accessories,different styles and styles are popular with young people at home and abroad,and their quality will naturally attract consumers’ attention.In the face of the growing market demand,manufacturers’ requirements for the size measurement and appearance defect detection of watch straps have also been gradually improved.The previous manual detection methods will seriously affect the production progress and sales of enterprises,especially the large impact of human subjective factors will easily lead to miscalculation,with low accuracy.Therefore,this paper designed a machine vision based watchband size measurement and defect detection system according to the production standards of enterprises,aiming at the size measurement and defect detection required in the production process of watchband.Firstly,this paper introduces the appearance characteristics and detection requirements of the watch band,analyzes the difficulties of size measurement and defect detection in the production process,and designs the overall structure and detection process of the detection system.Considering the pixel resolution,distortion,depth of field and other factors,a suitable visual platform of light source,lens and camera is built to collect images.For the camera distortion,the imaging model is established,and Zhang Zhengyou calibration method is used for black and white lattice calibration.The calibration experiment and back projection method verify that the method has high accuracy and low relative error.In order to improve the robustness of the image and facilitate the user’s observation in the image preprocessing stage,the advantages and disadvantages of filtering noise reduction,edge extraction and other algorithms are analyzed.After experimental comparison,it is found that median filtering noise reduction and Sobel operator edge extraction are better for subsequent image processing.In order to measure the size of the strap more efficiently,based on the shortcomings of the traditional template matching method,this paper adopts a composite image matching algorithm,which first encodes the pixel gray value to get the R-block,then performs rough matching according to the template and R-block,uses fine matching to solve the deviation problem caused by rough matching,combines the Hough transform to fit a straight line,and then uses the line coordinate information to get the result,by changing the brightness of the light source Change the position of the watch strap and inspect it in different computers to verify whether the size meets the requirements.In this paper,scratch defect,character defect and dirt defect are mainly detected.In order to ensure the position of real-time image is consistent with the standard image,ROI is selected and then affine transformation is used for image registration.The scratch defect is detected by top hat operation and maximum variance between classes;Compare the distance between the template and the contour point of the current image to determine whether characters are missing or overlapping;Dirt defect is detected by image gray difference.The shape of defects is extracted from Blob analysis,and the number,location,area and other information of defects are obtained by establishing the minimum bounding rectangle to classify and judge the defects.Finally,the strap size measurement algorithm studied in this paper is used to carry out experiments.By measuring the strap size under different conditions,the standard deviation of the measured length in the experimental data is less than 0.01 mm,and the standard deviation of the measured angle is less than 0.05 °.The appearance defect detection algorithm of watchband is used for verification,and the success rate of defect detection of multiple images reaches 97%.It can be concluded that the size measurement accuracy is high and the defect detection accuracy is high,meeting the production requirements of enterprises for watch straps.Figure [101] Table[14] Reference[70]...