中文摘要：

为解决超声渡越衍射时差（Time of flight diffraction,TOFD）检测图像中缺陷识别的问题,分析检测图像的特点,研究图像自动分割的算法,其中包括图像预处理及图像分割。提出基于信号互相关算法的图像预处理方法,在此基础上,应用峰值搜索方法提取焊缝区域,利用带控制标记符的分水岭变换对预处理的图像进行分割,从而识别出缺陷目标。利用提出的图像自动分割方法分割不同的超声TOFD检测图像。研究表明,基于信号互相关算法的图像校正方法在一定程度上可抑制检测图像的畸变,焊缝区域图像的提取可减少图像分割过程的计算量,从局部极值的角度出发的带控制标记符的分水岭变换实现缺陷目标的分割。与基于阈值方法的图像分割结果相比,图像自动分割算法较好地解决了近表面缺陷的识别问题,同时提出的方法也可用于含多个缺陷的图像分割。

英文摘要：

In order to identify the defect in the testing image obtained by the ultrasonic time of flight diffraction（TOFD） method,by analyzing the feature of the testing image,the automatic algorithm for image segmentation is studied,including image preprocessing and segmentation.The image preprocessing method is presented,which is based on the signal cross-correlation algorithm.Then the extraction of weld region based on the searching extreme points of signals is expressed.After that the watershed transform with the control marker is used to segment the preprocessed image,thereby the defect is extracted.Different ultrasonic TOFD testing images are segmented by the presented automatic image segmentation method.The results show that the signal cross-correlation algorithm basically suppresses the pattern distortion of the image and the extraction of weld region can reduce the calculation time of segmentation.Compared with the segmentation result by the threshold method,the presented automatic image segmentation method solves the identification of defect which is close to the lateral wave and the method also can be used for multiple defects identification.