Research On Measurement And Defect Detection Of Latex Gloves Based On Machine Vision

According to the analysis and forecast report on the market prospect of China's latex industry,the output of latex gloves in China is expected to exceed 10 billion pairs in 2020.According to the national standards and production process requirements,each latex glove needs to be strongly inspected.The main contents of the strong inspection include: dimension measurement,surface defect detection,etc.At present,the quality inspection of latex gloves still depends on manual work.In order to realize the automatic detection for latex gloves,this thesis mainly carried out the following work.Firstly,according to the process flow of manual inspection for latex gloves,the thesis gave the system block diagram of the latex gloves detection device based on machine vision.According to the testing requirements of latex gloves,the camera and light source used in the collection link were selected by theoretical analysis,the system control logic was expounded,and the communication protocol was designed.Secondly,the thesis studied the image filtering algorithms.According to the comparison experiments of different algorithms,the filtering algorithm suitable for the detection system was selected,and the edge information was extracted by Canny improved edge extraction algorithm.By analyzing the characteristics of latex gloves,a double threshold minimum external rectangular method was proposed to separate the defects of latex gloves from the outer contour.The defect size measurement was realized by using the minimum external rectangular method,and the outer size measurement of latex gloves was realized by the combination of Hough circle detection method and mathematical method.Thirdly,because of the difference of the surface defect types of latex gloves,a multi-feature SVM method was used to identify and classify the defect types.According to the characteristics of the device,the thesis proposed the localization method based on pixel domain division to determine the location of latex glove defects.Finally,the thesis implemented the algorithm studied in the host computer,and established the human-computer interactive operation interface of the host computer for experimental verification.The experimental results show that the accuracy of measuring and defect recognition of latex gloves is basically up to the expected goal.