产品CAD模型与CT切片模型的配准研究

【作者】 刘晶；

【导师】 张定华；

【作者基本信息】 西北工业大学 ， 航空宇航制造工程， 2003， 硕士

【摘要】 利用三维CT设备对工业产品进行无损检测时，不可避免地会存在定位偏差问题，从而要进行CT切片模型到CAD模型之间的配准。配准问题是图形图像处理领域研究的重要内容之一。目前国内大部分的研究是针对二维配准的研究，在工业方面针对三维配准的研究还不是很多。本论文针对模型的配准问题，分别从灰度信息和图像轮廓两个方面考虑，提出了基于灰度残差平方和的配准方法和基于惯量椭圆的配准方法。本论文在Microsoft Visual C++以及VTK环境下，进行程序的开发，通过仿真实验验证了两种算法的正确性和可行性，并把求得的配准参数同理论值进行比较，取得了比较满意的结果。主要完成的研究内容如下： ·匹配过程中的空间变换和重新采样。在三维模型匹配的过程中，不可避免地要做空间变换及重新采样，通过平移矩阵和旋转矩阵来建立起倾斜的CT切片模型到CAD模型之间的变换关系。在模型匹配过程中，为了得到变换后的CT切片，需要对图像进行重新采样。本论文采用二线性插值与三线性插值相结合的方法来进行重新采样。 ·基于灰度残差平方和的匹配算法。此方法首先通过图像模型的一阶矩，进行灰度重心匹配；然后将产品的图像模型与其CAD理论模型之间的残差平方和定义为优化目标函数，采用黄金分割方法搜索其极小值；同时利用产品上的结构特征，求出空间变换的位置匹配参数和角度匹配参数，最后根据匹配参数精确地建立起从产品CT切片模型到其CAD理论模型的空间变换。 ·利用椭圆匹配法和灰度重心相结合来进行匹配。此方法首先对CT切片图像提取轮廓点，然后把提取出的轮廓作为目标，接着借助转动惯量计算目标的惯量椭圆，并进一步对惯量椭圆匹配以获得所需的空间变换的参数，从而实现产品CT模型与理论模型之间的配准。 ·通过仿真实验验证上述两种算法。针对三维模型的配准问题，在Microsoft Visual C++以及VTK环境下，进行程序设计开发，实现上述匹配算法。并把配准后的三维CT模型同CAD理论模型的XY方向截面和YZ方向截面进行比较，验证算法的正确性和可行性。更多还原

【Abstract】 There are some position and orientation problems when a 3D CT equipment undertakes nondestructive testing to industrial products. So product’s CT slice model and its CAD model should be registered. Registration is an important problem in the field of graph and image processing. Now in our country a large amount of study is about planar registration study, three-dimensional registration study is few in industry field. For the registration of model, this paper takes into account gray information and image contour respectively. Registration method based on gray’s deviation square sum and registration method based on inertia ellipse are put forward. In Microsoft Visual C++ and VTK’s environment, the programming verifies the two algorithms by simulation experiment. Then evaluating registration parameters and theory values are compared, satisfactory results are received. The detail research works are as follows:· spatial transformation and re-sampling in registration. Spatial transform and re-sampling are carried out in three-dimensional model registration. Spatial transformation from skew CT slice model to its CAD model is established by transfer matrix and rotation matrix. In model’s registration, image should re-sample to get transformed CT slice. This paper adopts bilinear interpolation and tri-linear interpolation to re-sample.· algorithm based on gray’s deviation square sum. First the method proceeds to match gray centroid using one-order moment of image model. Then deviation square sum between product’s image model and CAD model is defined as optimized object function. We search the minimum by the golden section method and make use of specified structure character to get the position and orientation registration parameters of spatial transformation. Finally spatial transformation from product’s CT slice model to its CAD model is established according to registration parameters.· registration by ellipse matching and gray centroid. First this method extracts the contour points of CT slice images, then chooses extracted contour points as object. We can build up the inertia ellipses with the help of the moment of inertia and match these ellipses to obtain the parameters of the required spatial transformation. Thus registration between product’s CT model and theory model is realized.· verifying two kinds of algorithms above by simulation experiments. For the registration of three-dimensional model, the programming realizes the two algorithms above in the environment of Microsoft Visual C++ and VTK. Then XY section image and YZ section image between registered three-dimensional CT model and CAD model are compared to verify validity and feasibility of algorithms.更多还原