Design Of Defect Detection System For Cosmetic Cotton Roll

Cosmetic cotton roll is the raw material used in the production of cosmetic cotton sheet by cosmetic cotton cutting machine.The existing automatic cutting machine can not identify and remove defective fabrics.In actual production,it is necessary to detect the cotton roll quality in the previous process and make a mark in order to remove waste materials in the automatic cutting process.At present,the defect detection process is to detect the cotton roll defects through the naked eye of the inspectors,which has the problem of low detection efficiency and accuracy.Therefore,in this paper,an automatic detection system is designed for the defects of cosmetic cotton roll.Taking metal chips,stains,creases,spurs and tears as the detection objects,the overall scheme of the cosmetic cotton roll defect detection system is determined,the defect detection algorithm of cosmetic cotton roll is designed and tested,and the development of defect detection prototype software is completed.The research contents of this paper are as follows（1）The whole scheme of cosmetic cotton roll defect detection system is designed.Combined with the design requirements of the detection system,the overall architecture of the detection system is designed based on the existing detection equipment.After summing up the various components of the image acquisition system,the corresponding hardware is calculated and selected according to the detection requirements,and the design of the image acquisition system and processing system is completed.（2）An algorithm for detecting cotton roll defects is designed.Firstly,the defect region is extracted,a variety of defect extraction algorithms are compared and studied,and a defect extraction method with different processing according to different defects is designed.This method can extract the defect information in the image efficiently and completely.Then,by analyzing the samples of various types of defects,a scheme to determine the types of defects through the characteristic parameters of defects is determined.After comparing Gaussian mixture model（GMM）and multilayer perceptron model（MLP）,Gaussian mixture model classification algorithm is selected.The experimental results show that published algorithm accuracy reaches 98.07%,and that speed reaches 73.39m/min.,which meets the detection requirements of cosmetic cotton rolls.（3）The software prototype system for detecting cotton roll defects is developed.The defect detection algorithm proposed in this paper is integrated by using Halcon and c# hybrid programming,after logging in,the user can set the camera parameters and control the camera to collect images,grade the cotton roll after detecting defects,use the database as the center for storing defect data and cotton roll data,and have an intuitive and friendly human-computer interaction page.