基于RBF网络的焊缝缺陷图像的识别与诊断技术研究  

Research on Recognition and Diagnosis Technology of Weld Defect Image Based on RBF Network

作　　者：乔荣华[1] 董建刚[1] 张永恒[1] 

机构地区：[1]榆林学院信息工程学院,陕西榆林719000

出　　处：《热加工工艺》2016年第1期217-220,共4页Hot Working Technology

基　　金：陕西省教育厅科学研究计划项目(2013JK1168)

摘　　要：本文主要研究了采用中值滤波、LOG算子的边缘检测以及DSA减影等数字图像处理技术对现有的焊缝缺陷图像进行处理,获得带有各种焊缝缺陷特征的大量图像样本,并用RBF神经网络对样本进行训练,最终获得具有较高识别精度的RBF网络。利用该网络可以大大提高对焊缝缺陷的判断效率,具有一定实用意义。How to deal with the processing of weld defect was studied by using digital image processing techniques, such as image with the median filtering, LOG edge detection operator and DSA subtraction, and a large number of image samples with various weld defect characteristics were obtained. Then the samples were trained by RBF neural network, and the final RBF network was achieved which had a high enough recognition accuracy. The use of the network can greatly improve the efficiency of weld defect judgment, so the research has a certain practical significance.

关 键 词：图像处理技术 焊缝缺陷识别 RBF神经网络 

分 类 号：TG441.7[金属学及工艺—焊接]