基于ICT切片图像相邻层信息的缺陷自动识别方法

Automated defect recognition method based on neighbor layer slice images of ICT

现有的工业计算机断层成像(ICT)图像缺陷识方法中,多采用对单张图像进行孤立评判方法,此类方法未能考虑到单张图像在相邻层图像信息关联性,因而易将孤立的噪音视为缺陷,造成误判。为解决这一问题,提出一种基于序列ICT切片图像自动识别方法,该方法将识别过程分为两步:单张图像的潜在缺陷提取和相邻层图像缺陷的匹配。第一步运用传统方法识别出每张图像中所有潜在缺陷;第二步根据真缺陷在相邻层具有匹配关系而伪缺陷则相对孤立的特点,通过分步匹配的方法确定每张图像上所有潜在缺陷在相邻层图像上的匹配关系,区分出真伪缺陷。最后通过实例验证表明:利用该方法可以有效得提高真缺陷得识别率,降低误判率。

The current automated defect recognition of industrial computerized tomography（ICT） slice images is mostly carried out in individual image. Certain false detections would exist for some isolated noises can not be wiped off without considering the information of neighbor layer images. To solve this problem, a new automated defect recognition method is presented based on a two-step analysis of consecutive slice images. First, all potential defects are segmented using a classic method in each image. Second, real defects and false defects are recognized by all potential defect matching of neighbor layer images in two steps based on the continuity of real defects characteristic and the non-continuity of false defects between the neighbor images. The method is verified by experiments and results prove that the real defects can be detected with high probability and false detections can be reduced effectively.