高压涡轮叶片工业CT图像轮廓提取方法研究  

Research on Contour Extraction Method of High Pressure Turbine Blade Industrial CT Image

作　　者：曾凤英 韩晨垚 姚振文 ZENG Fengying;HAN Chenyao;YAO Zhenwen(China Gas Turbine Establishment,Aero Engine Corporation of China,Mianyang 621000,China;Xi'an Jiaotong University,Xi'an 710049,China)

机构地区：[1]中国航发四川涡轮研究院,四川绵阳621000 [2]西安交通大学,西安710049

出　　处：《无损探伤》2024年第3期12-16,20,共6页Nondestructive Testing Technology

摘　　要：针对高压涡轮叶片CT扫描图像轮廓不清晰的问题,提出首先采用基于卷积神经网络模型DexiNed对断层扫描图像进行轮廓的粗提取,然后结合形态学运算方法中的开运算、轮廓细化和轮廓去毛刺方法对提取出的轮廓进行后处理。实验结果表明该方法可以非常准确地提取出高压涡轮叶片断层扫描图像中的结构轮廓,为高压涡轮叶片内部结构尺寸的准确测量提供基础,并具有可推广适用性。In view of the problem that the contour of CT scan images of high-pressure turbine blades is unclear,DexiNed based on convolutional neural network model is proposed to rough extract the contour of the tomography images,and then the extracted contour is post-processed by combining the open operation,contour thinning and contour deburring methods in the morphology operation method.The experimental results show that the proposed method can extract the structure profile from the CT image of high-pressure turbine blades very accurately,which provides a basis for the accurate measurement of the internal structure size of high-pressure turbine blades,and can be generalized.

关 键 词：CT图像 轮廓提取 涡轮叶片 卷积神经网络 

分 类 号：TG115.28[金属学及工艺—物理冶金]