作者：田密1,2, 韩军1, 吴飞斌1, 龙晋桓1, 黄惠玲1
作者单位：1. 中国科学院福建物质结构研究所，福建 福州 350002;
2. 中北大学电气与控制工程学院，山西 太原 030051
关键词：超声C扫描成像;声束重叠;区域描绘子;纹理测度;缺陷分割
摘要：
超声C扫描成像中经常存在成像对比度低边缘模糊不清等现象，从而导致图像分割不完整。针对该问题，提出基于统计法纹理描绘子的图像分割算法。基于单晶探头C扫描成像中采样点之间声束重叠的特点，首先进行超声声束分析并且选取声束重叠数据；然后采用区域描绘子的相关纹理测度替代传统的幅值特征值，对采样区域中缺陷的特征信息进行描述，从而提高超声成像分辨率和对比度；最后采用数学形态学进行滤波，进一步提高缺陷的识别度。通过304不锈钢标准试块底面圆形盲孔C扫成像试验进行验证并与自适应阈值分割算法比较，实验结果显示该文算法可以得到清晰完整的缺陷分割区域，并且比自适应阈值分割算法更为有效可靠，尤其对于易忽略的小尺寸缺陷同样得到良好的分割效果。
Ultrasonic image segmentation algorithm based on statistical texture descriptors
TIAN Mi1,2, HAN Jun1, WU Feibin1, LONG Jinhuan1, HUANG Huiling1
1. Fujian Institute of Research on the Structure of Matter, Chinese Academy of Sciences, Fujian 350002, China;
2. School of Electrical and Control Engineering, North University of China, Taiyuan 030051, China
Abstract: In the ultrasonic C-scan imaging, there are many problems such as low contrast and blurred edge, which lead to incomplete image segmentation. To solve this problem, an image segmentation algorithm based on statistical texture descriptors is proposed. Based on the overlapping characteristic of sound beams between sampling points in single crystal probe C-scan imaging, through the feature analysis of ultrasonic beams, the sound beams overlapping data are selected from each subwindow of the ultrasonic image. And then the characteristic information of the defect in the sampling area is described by the correlation texture measure value of the area descriptors which substitute for the traditional amplitude characteristic value, so as to increase the resolution and contrast of ultrasonic imaging. Finally, the defect recognition rate is further improved by digital morphological noise reduction method. In this paper, the algorithm is verified by the instance of ultrasonic C-scan imaging of the 304 stainless steel standard with circular blind holes on the bottom of it. The segmentation results of the proposed algorithm and the traditional adaptive threshold algorithm are compared quantitatively. Experimental results show that the algorithm in this paper can obtain clear and complete defect segmentation regions, and is more accurate and reliable than the adaptive threshold segmentation algorithm, especially for small size defects which are easy to be neglected, satisfied result of defection region segmentation are also obtained.
Keywords: ultrasonic C-scan imaging;sound beam overlapping;region descriptors;texture measurement;defect segmentation
2020, 46(8):121-125 收稿日期: 2020-02-11;收到修改稿日期: 2020-03-17
基金项目: 国家自然科学基金（11705208）；中国科学院对外重点合作项目（121835KYSB20180062）；福建省科技计划项目（2018T3007，2019T3020）；泉州市科技计划项目（2018C105R，2019C097R）
作者简介: 田密（1993-），男，山东潍坊市人，硕士，主要从事超声检测及超声无损评价评估研究
参考文献
[1] 李灼华. 超声无损检测技术的现状和发展趋势[J]. 硅谷, 2013, 6(21): 11-16
[2] GU Z, LI X, HUANG H, et al. Deep learning-based image segmentation on multi-modal medical imaging[J]. IEEE Transactions on Radiation and Plasma Medical Sciences, 2019: 1-1
[3] MA Z Q, WANG Y S, SONG Z B, et al. The color tread image segmentation based on improved labeled watershed[J]. Journal of Graphics, 2018, 39(1): 36-42
[4] SHENG Z. Segmentation of ultrasonic TOFD testing image based on watershed method[J]. Journal of Mechanical Engineering, 2011, 47(8): 35
[5] 刘博, 黄剑华, 唐降龙, 等. 结合全局概率密度差异与局部灰度拟合的超声图像分割[J]. 自动化学报, 2010, 36(7): 951-959
[6] CHAIBOU M S, CONZE P H, KALTI K, et al. Adaptive strategy for super pixel-based region-growing image segmentation[J]. Journal of Electronic Imaging, 2018, 26(6): 1
[7] 李雄兵, 周晓军, 吴思源, 等. 超声C扫描图像缺陷标记及边缘跟踪的研究[J]. 传感技术学报, 2006(6): 326-329
[8] 任明照, 高东海, 郑韵娴. 铁路车轮超声波B型检测的图像分割算法[J]. 无损检测, 2017(7): 12-15
[9] 倪豪, 郑慧峰, 王月兵, 等. 基于自动种子区域生长的超声B图像缺陷分割方法[J]. 计量学报, 2018, 39(6): 128-133
[10] 李俊山, 李旭辉. 数字图像处理[M]. 2版. 北京: 清华大学出版社, 2013.
[11] 刘辉, 张云生, 张印辉, 等. 基于灰度差分统计的火焰图像纹理特征提取[J]. 控制工程, 2013, 20(2): 213-218
[12] LORIS N, SHERYL, ALESSANDRA L. Random interest regions for object recognition based on texture descriptors and bag of features[J]. Expert Systems with Applications, 2012, 39(1): 973-977
[13] LI Z, YANG Y, JIANG W. Multi-scale morphologic tracking approach for edge detection[C] // International Conference on Image and Graphics. San Antonio, IEEE, 2007: 358-362.
[14] JU Z, ZHOU J, XIAN W, et al. Image segmentation based on adaptive threshold edge detection and mean shift[C] // International Conference on Software Engineering & Service Science. IEEE, 2013.
[15] 杜磊, 李立轻, 汪军, 等. 几种基于图像自适应阈值分割的织物疵点检测方法比较[J]. 纺织学报, 2014, 35(6): 56
[16] 盛恒, 黄铭, 杨晶晶. 基于Faster R-CNN和IOU优化的实验室人数统计与管理系统[J]. 计算机应用, 2019, 39(6): 1669-1674
[17] 谢行雨, 王玲. 基于纹理相似栈的超声图像分割方法[J]. 计算机工程, 2019, 45(2): 240-244
