采用特征向量和神经网络的管缺陷检测算法  

An Algorithm for Flaw Detection Using Feature Vectors and Neural Networks

作　　者：郑晓玲[1] ZHENG Xiaoling(School of Intelligent Manufacturing, Liming Vocational University, Quanzhou 362000, China)

机构地区：[1]黎明职业大学智能制造学院,福建泉州362000

出　　处：《黎明职业大学学报》2020年第3期86-91,共6页Journal of LiMing Vocational University

摘　　要：利用机器视觉技术自动检测管表面缺陷存在缺少数据集,环境中照明条件不稳定等问题,导致管表面缺陷自动检测工作难度大,准确率低。提出一种在有限数据集和不稳定光照条件下的管缺陷检测算法,利用特征向量和神经网络技术对缺陷进行检测和分类。试验表明:对缺陷分类的准确率平均达到95%。At present,the automatic detection of tube surface defects using machine vision technology are suffering from problems such as the lack of data sets,unstable lighting conditions in surroundings,etc.,which lead to difficulties and low accuracy of automatic detection of tube surface defects.An algorithm under the conditions of finite data sets and unstable illumination was then proposed with the use of feature vector and neural network technology to detect and classify defects.Experimental results showed that the average accuracy of defect classification reaches as high as 95%on average.

关 键 词：机器视觉 表面缺陷检测 神经网络 

分 类 号：TP249[自动化与计算机技术—检测技术与自动化装置]