摘要
提出并设计了一种基于机器视觉的注塑件焦料杂质缺陷检测系统,通过对多光源非线性自动调光算法的研究,提高了复杂生产环境下注塑件图像采集质量,并经高斯滤波、灰度图像、自适应阈值处理、Canny边缘检测等图像处理技术,可实现注塑件不同检测精度下缺陷检测。经现场测试证明,在1 mm以上缺陷检测精度下,杂质缺陷的召回率为0.918,误检率为0.046,识别速度为每秒30帧,图像处理算法性能稳定,解决了复杂注塑件生产环境下检测精度低、检测效率低的问题。
In this paper,a machine vision-based defect-detecting system for coke impurities in the injection molding parts was proposed and designed.Through the study of the multi-light source nonlinear auto-dimming algorithm,high-quality images of the injection molding parts were collected in a complex production environment and through image processing technologies such as Gaussian filtering,gray image and adaptive threshold processing,as well as Canny edge detection,the injection molding parts can be detected at different precision.The field test shows that,as for a defect detection accuracy of more than 1 mm,the recall rate of the parts with impurity defects is 0.918,the false detection rate is 0.046 and the recognition speed is 30 fps.The image processing algorithm has stable performance,which solves both low detection accuracy and efficiency in complex injection molding production.
作者
王亓才
冯小辉
史卜凡
冀国正
曹怀祥
袁涛
黄元凤
WANG Qi-cai;FENG Xiao-hui;SHI Bu-fan;JI Guo-zheng;CAO Huai-xiang;YUAN Tao;HUANG Yuan-feng(Shandong Giant E-Tech Co.,Ltd.)
出处
《化工自动化及仪表》
CAS
2024年第1期113-119,共7页
Control and Instruments in Chemical Industry
基金
山东省重点研发计划(重大科技创新工程项目)(批准号:2020CXGC011201)资助的课题。
关键词
机器视觉
注塑件缺陷检测
图像处理
视觉控制平台
machine vision
defect detection of injection molding parts
image processing
visual control platform
