Multisensor Integration and Discrete Geometry Processing for Coordinate Metrology

Multisensor integration systems are used to fulfill the increasing requirements on the complexity andaccuracy of coordinate metrology. This thesis investigates on the achievement of multisensor integration incoordinate metrology from two main parts: function and data modeling; and discrete geometry processing.Function and data modeling allow obtaining clear understanding of the system and guiding the systemdevelopment. Hierarchical function models of the system are built using IDEF0 methodology. An ontologybased method is employed to model the various data and their relationships in the system with structuredrepresentation. Efficient methods for discrete geometry processing to support multisensor integration are themain focus of this dissertation. Reliable estimation of the curvatures is essential for discrete shapesprocessing. A tensor based method is proposed for discrete curvature estimation. The proposed methodprovides better performance according to a comparative analysis. Registration is one of the most importantand decisive steps of multisensor integration. Registration algorithms enable to align the different point datacaptured in the respective sensor’s coordinate systems. A registration method is proposed to align differentdiscrete shapes with unknown correspondences. A combination of curvature information and Euclideandistance is used to improve the registration performance. A new method is developed for shape recognitionand segmentation based on shape index and curvedness. The local surface type of a vertex is recognizedaccording to its shape index and curvedness. The vertices are then clustered according to their local surfaces.A connected region generation algorithm is then implemented to generate the segmentation results.Considerable cases are tested and the method provides satisfying results. A new software frameworkDSP-COMS is developed as the testing platform of the developed methods. A detailed case study of anindustrial workpiece digitized by laser scanning and touch probing is presented. The measured point data arethen processed using DSP-COMS. The results show that the proposed methods can provide satisfying resultsfor the studied case.