【作者】 刘鹏鑫；

【导师】 王扬；

【作者基本信息】 哈尔滨工业大学 ， 机械制造及其自动化， 2010， 博士

【摘要】 光学扫描系统和三坐标测量机(CMM)是逆向工程中广泛使用的数字化设备,二者的组合使用满足了逆向工程中对实物数字化的高精度、高效率的应用需求。多传感组合测量系统的出现,也为这种组合提供了硬件基础和支持。现有的逆向建模方法主要还是针对同密度、同精度的同类测量数据的。为了满足现代数字化系统对逆向工程模型重建方法提出的新要求,弥补现有逆向工程建模方法和软件的不足,本文研究了基于光学扫描和CMM测量数据的模型重建方法及其关键技术。首先,实验分析了光学扫描系统和CMM的优缺点及其对模型重建的影响,通过测量实例证明了二者的互补性,进而提出了基于光学扫描和CMM测量数据的模型重建方法,并详细论述了该方法的框架、研究内容和关键技术,为后续的深入研究提供了指导方向。以三角网格模型表达了光学扫描获取的数据,研究了三角网格模型的微分几何特性估算方法。提出的曲率估算方法能够识别Voronoi曲率估算方法的估算异常区域,以异常区域内的每个顶点的2环邻域作为k邻域,采用加权的局部抛物面拟合法对异常区域曲率进行了修正。虽然算法精度位于Voronoi法和抛物面拟合法之间,但是运行效率高于局部抛物面拟合法,而且由于局部抛物面拟合的引入使该算法不易受三角网格形状的影响,并具有一定的抑噪能力。为了能够清晰的识别网格模型的凹凸特征和曲率估算异常区域,基于颜色索引表和直方图均衡原理实现了曲率的可视化。以法矢为度量准则,提出了一种分步式的、噪声鲁棒的网格分割方法。基于均值漂移理论对网格法矢进行了滤波,选取了14个单位向量作为聚类中心,应用K-means算法对网格法矢进行聚类,根据聚类的结果采用区域生长法实现了网格初始分片,最后以面片的曲度为标准,对面片进行了融合以识别规则曲面,这些面是与CMM测量特征对应的潜在区域。对CMM测量数据和光学扫描数据的配准进行了研究。对于不存在光学扫描盲区的配准,提出了一种不依靠简单几何体作为基准的配准算法。通过近似选取3对对应点,求解坐标变换矩阵实现初始配准;然后对光学扫描点云进行三角剖分,以CMM测量所获点集向对应三角片的投影寻找对应点,之后进行坐标变换迭代求解,从而实现了不存在对应点的点集之间的配准。对于存在光学扫描盲区的配准,给出了一种基于配准工作台的配准方法,以基准点连接线段的中点代替原来的基准点改进了三点定位算法,实验对比了几种配准方法的误差,证实了该方法的可行性。在网格分割和配准的基础上,实现了规则曲面区域的数据融合以剔除与CMM测量特征重叠的三角网格。针对自由曲面区域的数据融合问题,提出了一种自适应的数据融合方法,根据CMM的测量特征实现了对三角网格模型的自适应分割,并基于夹角准则识别了融合区域。因为无需重新整体扫描而缩短了数字化时间,所以该方法对于三角网格模型的快速定制、变形及基于原有实物样件的变形设计、自由曲面设计及快速原形制造等都很有意义。最后给出了规则曲面和自由曲面重建方法,根据模型优化重建策略实现了基于光学扫描和CMM测量数据的模型重建。更多还原

【Abstract】 Optical scanning system and three coordinate measuring machines (CMMs) are digitization equipments widely used in reverse engineering. The combined use of both satisfies the application requirements of reverse engineering for high accurate and high efficient digitization. The appearance of multi-sensor combined measuring system also provides the hardware foundation and support for the combination. The exising model construction methods in reverse engineering are mainly based on the similar data which has the same density and accuracy. In order to satisfy the new requirement of the modern digtitization system for model construction method and make up the deficiency of model construction method and software, the method and key technologies of model reconstruction from optical scanning and CMM data are researched in this thesis.Firstly, the strengths and weaknesses of optical scanning system and CMM and their effect for model reconstruction were analyzed by experimentions. Study cases proved these two ditigition methods were complementary. Therefore, method of model reconstruction from optical scanning and CMM data were provided. This thesis also gave a detailed description about the framework, research contents and key technologies which offer the direction for the further study.The optical scanning data were expressed by triangular mesh model. Estimation on differential geometric properties of triangular mesh model was studied. The provided method of curvature estimation can identify the potential anomalous regions of Voronoi method through choose the 2-ring neighbor of each vertex in these regions as k-ring neightbor, using local paraboloid fitting corrected the curvature. Although the accuracy of the provided method was between the Voronoi and local paraboloid fitting, its operation efficiency was higher than local paraboloid fitting. This method was not easy to be effected by triangle shape and had some ability to suppress noises due to introducing of the local paraboloid fitting. Curvature visualization was also realized based color index and histogram equalization algorithm in order to clearly identify the concave or covex features and anomalous regions of curvature estimation.A step and robust segmentation method was provided using normal as a metric. Firstly, mesh normals were filtering through the extension of mean shift. Then a K-means algorithm according to the normal vectors was used by means of chosen fourteen directions as clustering centers. After that initial segmentation was completed using region grow algorithm. Patches merging was the last step, using curvedness as a metric to recognize regular surfaces which are potential regions corresponding with CMM measuring features.Registration between CMM and optical scanning data was researched. In the case of without optical scanning blind areas, a registration method was proposed, in which primitive artifacts were not required as datum, three pairs of points were selected to compute a transform matrix for initial registration. Afterwards, in order to get corresponding points, point set from CMM measurement was projected to nearest triangular mesh constructed from optical scanning data, and the transformation could be achieved by iterative method. This method realized the registration of point sets of non-corresponding points. When having optical scanning blind areas, a registration method using work table was presented, and improved three points location algorithm was used through the middle point of the line segment connected two datum marks instead of initial datum marks. Registration errors of several registration methods were compared for further confirm the feasibility of this improved method.Data merging at regular surfaces based mesh segmentation and registration was realized in order to delete the triangles overlapping with CMM measuring data. According to the data merging at free-form surfaces, this thesis presents an adaptive data merging method which was completed by adaptive mesh segmentation based CMM measuring features and identification of merging regions based on the angle condition. The method would enable to create the customized model and deformation because it might shorten the additional scanning time. It will be practically useful for the surface modification based on the physical part changes, free-form surface design and rapid prototyping manufacturing. The reconstruction methods of regular suface and free-form surface have been given, and then realize the model reconstruction from optical scanning and CMM data instructed by strategy for optimization model reconstruction.更多还原