Detection and classification of weld surface defects based on laser vision sensor

DING Xiaodong; LI Yangjin; GAO Xiangdong; ZHANG Yanxi; YOU Deyong; ZHANG Nanfeng

Views : 0 下载量: 3145 CSCD: 8

PDF
Export
Share
Collection
Album

Detection and classification of weld surface defects based on laser vision sensor
Vol. 49, Issue 7, Pages: 78-83(2019)
Published：2019
DOI：
稿件说明：

移动端阅览

DOI：

DING Xiaodong, LI Yangjin, GAO Xiangdong, et al. Detection and classification of weld surface defects based on laser vision sensor[J]. Eletric Welding Machine2019, 49(7): 78-83. DOI：

摘要

为实现焊缝表面质量的自动检测

研究线激光视觉传感的焊缝表面质量检测方法

分析焊缝表面缺陷特征提取算法及焊缝缺陷分类模型。针对焊缝表面中的凹陷、咬边和气孔等缺陷

分析不同类型缺陷在焊缝激光条纹图像中的几何形态及空间分布特点

并结合斜率截距法与分段区间检测法提取焊缝表面缺陷的特征点。利用特征提取方法识别焊缝表面缺陷的7个特征参数

设计基于三层BP神经网络的焊缝缺陷分类模型

将提取的缺陷特征作为网络的特征输入进行网络训练。试验结果表明

所建立的焊缝缺陷分类模型可识别凹陷、咬边、气孔等焊缝表面缺陷

整体识别率达91.51%。

Abstract

In order to realize the automatic detection of weld surface defects

a weld surface quality detection method based on line laser vision sensor is studied

the extraction algorithm of weld surface defect characteristics and a classification model of weld defects are analyzed. For the defects such as pit

undercut and blowhole in the weld surface

the geometric shape and spatial distribution characteristics of different defects in laser stripe images are analyzed. Also

the feature points of weld defects are extracted by slope intercept method and segmented interval detection method. Seven characteristic parameters of weld defects are detected by feature extraction method to design a weld defect classification model based on three-layer BP neural network. These extracted seven defect characteristics are input for network training as the network characteristics. The results show that the proposed method can detect the weld surface defects such as pit

undercut and gas pore

and the overall recognition rate of the classification model can reach 91.51%.