基于3-D Facet模型的亚体素边缘检测算法研究  

A Study of Subvoxel Edge Detection Method Based on 3-D Facet Model

作　　者：王凯[1] 张定华[1] 黄鹤龄[1] 毛海鹏[1] 梁亮[1] 

机构地区：[1]西北工业大学,西安710072

出　　处：《机械科学与技术》2005年第7期865-868,共4页Mechanical Science and Technology for Aerospace Engineering

基　　金：国家自然科学基金项目(50375126);航空科学基金项目(00H53076)资助

摘　　要：在产品的虚拟测量中,需要获得产品内外表面的3D高精度轮廓信息。本文首先介绍了一种直接从CT切片序列获得表面点的方法,通过3DFacet模型计算方向导数并根据沿梯度方向的二阶导数零点获得边缘点的亚体级精确位置。该方法可以检测到法向垂直于切片的表面,并且由于考虑了待检测切片相邻层的信息,与仅考虑单层信息的2D亚像素边缘检测方法相比,它抗噪能力更强、精度更高。其次,本文提出了一种改进方法,通过缩减候选像素的预处理,在很大程度上提高了原始算法的处理速度。最后将改进算法应用到航空发动机叶片仿真模型的检测中,结果表明本算法检测精度可达0.1个像素左右,速度较原始算法提高约10倍。For product virtual measurement application, three-dimensional (3-D) internal and external surfaces information in high accuracy of the object is required. At first, a method for detecting surface points directly from CT image sequence is introduced, which obtains the subvoxel accuracy by using facet model to estimate the local 3-D directional derivatives and location of the zeros of the second 3-D directional derivatives along the direction of gradient. Using this strategy, surfaces whose normal vectors are vertical to the slices can be detected. Furthermore, considering the information of slices adjacent to the being processed one, the method can reduce more noises and achieve higher accuracy when compared with the 2-D subpixel edge detectors which only use single slice information. Then we present an improved algorithm in this paper which can accelerate the original one greatly by preprocessing of reducing the number of edge pixel cand: dates. Finally, as an illustration, we apply the method to an aeroengine blade simulated model and the results show that about 0.1pixel accuracy can be reached and the speed can be increased about 10 times.

关 键 词：工业CT 边缘检测 亚像素级精度 亚体素级精度 FACET模型 

分 类 号：TG8[金属学及工艺—公差测量技术]