机电工程技术2024，Vol.53Issue(7)：153-157,5.DOI:10.3969/j.issn.1009-9492.2024.07.032
基于Halcon深度学习的高分子薄膜表面缺陷检测
Surface Defect Detection of Polymer Film Based on Halcon Deep Learning
摘要
Abstract
The types and characteristics of various defects on the surface of polymer films,as well as the potential damage they may cause to the performance and functionality of polymer films are elaborated.The aim is to demonstrate the necessity of detecting surface defects in polymer films during the production process.In response to the problems of low detection accuracy caused by the small number of samples in various defect categories and insufficient changes in defect features in the application of deep learning large model training in surface defect detection of polymer thin films,a global context outlier detection algorithm based on Halcon is applied,and the GC-AD Combined training model of the detection algorithm is elaborated.It is a network structure that integrates Faster RCNN and Auto Encoder.On the basis of relatively limited qualified images and image data containing defects of polymer films,an experimental design is carried out using the GC-AD Combined network model to train polymer film image samples.After the training completed,multiple sets of film image data are subjected to inference testing.At the same time,comparative experiments are designed with network models such as CNN,Mobile-net and VGG for detection.The experimental results show that the average accuracy of the GC-AD Combined network model used is as high as 98.07%,which can meet the surface defect detection task of blow molded films.The detection accuracy is significantly better than the other three network models,further demonstrating the effectiveness and superiority of using the GC-AD Combined model based on the Halcon global upper and lower anomaly detection algorithm to complete the surface quality detection task of polymer films.关键词
深度学习/Halcon/图像处理/缺陷检测/机器视觉Key words
deep learning/Halcon/image processing/defect detection/machine vision分类
计算机与自动化引用本文复制引用
邓海云,陈新辉..基于Halcon深度学习的高分子薄膜表面缺陷检测[J].机电工程技术,2024,53(7):153-157,5.基金项目
广东省科技专项资金项目(STKJ2023056) （STKJ2023056）
