Machine Vision For Hand Mold Surface Defect Detection

In recent years,nitrile gloves in China is in the stage of rapid growth cycle,in the popularity and application of disposable gloves,China and overseas there is a significant gap in the rapid growth stage,the market demand will continue to grow with the development of medical,food,chemical and other fields,nitrile gloves market prospects are also very broad.At present,the quality inspection of hand mold mostly relies on manual inspection,which is the traditional method of defect detection,the accuracy and reliability of which depends on personnel skills and experience.Although manual inspection can find some surface defects,it is not effective for the detection of hidden defects and minor defects.As production speeds continue to rise,manual inspection cannot meet the high standards and real-time requirements of the nitrile glove manufacturing process.During glove manufacturing,the surface of the hand mold inevitably rubs against other production tools,resulting in defects such as staining,missing fingers and cracks on the surface of the hand mold,and it is these defects that cause quality problems in gloves.In this context,the hardware,image processing techniques,feature extraction methods and classification methods used to detect defects on the surface of hand molds are investigated in this paper.The main work is as follows:(1)Based on the basic principles of machine vision technology,this study explores the lighting method and the hardware setup of the image acquisition system for defect detection on the surface of hand mold.By analyzing the glove production process,hand mold defect types,forms and causes,and combining the performance of relevant equipment devices in the market,a hand mold surface image acquisition system based on the forward illumination method and the surface array CMOS camera shooting method was designed.After the experimental platform is built,the verification of experimental results shows that the design scheme basically meets the requirements of hand mold surface defect detection.(2)After acquiring the hand mold surface image,this study uses color space,connected domain detection algorithm and morphological algorithm to process the image based on the external features of the hand mold.In particular,by improving the morphological gradient algorithm and using the basic morphological operations,the edge contours and cracks obtained are clearer and have higher accuracy and robustness to noise compared with the traditional morphological gradient algorithm.This processing process initially completes the processing of hand mold surface defect images and lays the foundation for the subsequent implementation of hand mold surface defect classification.(3)In this study,a support vector machine(SVM)model based on the geometric features of defects with grayscale co-occurrence matrix texture features is designed for the classification of hand mold surface defects.The model fuses the geometric features of defects with grayscale covariance matrix features,and the principle of SVM classifier and model construction process are introduced.In addition,the accuracy of the grayscale covariance matrix in hand mold surface defect classification with different angle parameters and window sizes is compared,and the grayscale covariance matrix parameters with the highest accuracy are selected for feature extraction of hand mold surface defects.Finally,the SVM was used to classify the hand mold surface defects.