基于边缘检测的焊接RT图像中缺陷分割方法  

Method of Defect Segmentation in RT Weld Image Based on Edge Detection

作　　者：耿道华[1] 徐桂云[1] 李国庆[1] 王恩元[1] 

机构地区：[1]中国矿业大学

出　　处：《煤矿机械》2006年第10期77-79,共3页Coal Mine Machinery

基　　金：江苏省博士后科研基金(0502010B);中国矿业大学科技基金(2005B005)

摘　　要：针对RT焊接图像,提出一种sobel算子与动态的自适应阈值相结合的边缘检测方法,并利用基于奇偶判断的填充算法,准确提取缺陷。对RT焊接图像进行边缘检测分割前,为去除噪声、检测和识别特征更加容易,首先在图像预处理中去除焊缝外背景,并进行了全方位多级组合滤波、局部区域的直方图均衡处理,然后还利用列灰度波形分析划分焊缝内区域。Aiming at RT weld image, a method of edge detection which combines sobel operator and dynamic adaptive threshold is put forward. At the same time, the filling algorithm based on parity judgment is used to extract defect accurately. For eliminating noise, detecting and identifying the characteristic easily, the weld image should be preprocessed before the segmentation of edge detection to RT weld image. The preprocessing includes weld background removing, all - directional multi -class filtering, local histogram equalization and weld region dividing using wavelet analysis.

关 键 词：RT焊接图像 缺陷 边缘检测 分割 区域填充 

分 类 号：TH877[机械工程—仪器科学与技术] TG406[机械工程—精密仪器及机械]