基于分块分形的工业CT图像缺陷自动定位算法  

Automatic locate algorithm for the defects of industrial CT based on block fractal

作　　者：陈培兴[1,2] 王明泉[1,2] 李世虎[1,2] 侯慧玲[1,2] 王玉[1,2] 

机构地区：[1]中北大学仪器科学与动态测试教育部重点实验室,山西太原030051 [2]中北大学信息与通信工程学院,山西太原030051

出　　处：《电子技术应用》2015年第6期147-149,153,共4页Application of Electronic Technique

基　　金：国家自然科学基金(61171177);山西省青年科技研究基金(2012021011-1);山西省科学技术发展计划(工业)项目(20140321010-02)

摘　　要：针对传统的缺陷定位必须经过图像分割和缺陷提取等步骤,识别过程比较麻烦而且费时,提出了一种基于分块分形的工业CT图像缺陷自动定位算法。该方法首先对图像进行分块处理,对每个分块区域进行分形维数计算。通过分形维数频域分布直方图进行阈值处理,标记边缘块,最后通过连通区域处理标记块,进而对缺陷进行标记定位。通过对含有不同缺陷数目的固体火箭发动机模型工业CT图像处理,均可以准确地定位缺陷。实验结果表明,该方法能有效、准确地自动定位工业CT图像缺陷,且具有较强的鲁棒性。For traditional defect location must through steps such as image segmentation and defect extraction, identification process is cumbersome and time-consuming, so this paper proposes a automatic detection algorithm for the defects of industrial CT based on block fractal. To begin with, the method deal with the image into blocks and fractal dimension is calculated on each sub-block area. Then, it sets a threshold according to the frequency distribution histogram of fractal dimensions, and marks the edge of the block. Finally, by dealing with the connected region of marked blocks, it is able to locate and mark defects. Through the processing of solid rocket motor model industrial CT images which contain a different number defects, it can accurately locate defects. The experiments indicate that this method is effective and accurate on automatic locate defects of industrial CT image, and has a strong robustness.

关 键 词：缺陷定位 分形 工业CT 分块 区域连通 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]