Feature based shape interrogation, the process of extracting or describing features for a model surface, is a fundamental technique in Computer Aided Design/Manufacturing systems. This thesis focuses on the feature based shape interrogation methods related to sweep scanning path planning and part-in-whole matching.

Sweep scanning is a novel surface inspection technique with a much higher efficiency than the traditional ones. To dates, there are no published reports on its continuous path planning. In this thesis, a new type of feature, the curved reflection axis, is formulized for free-form surfaces to provide a powerful algebraic model for planning sweep scanning path. It is defined to be a smooth 3D curve, any point of which is the local reflection center of some surface points. The...[ Read more ]

Feature based shape interrogation, the process of extracting or describing features for a model surface, is a fundamental technique in Computer Aided Design/Manufacturing systems. This thesis focuses on the feature based shape interrogation methods related to sweep scanning path planning and part-in-whole matching.

Sweep scanning is a novel surface inspection technique with a much higher efficiency than the traditional ones. To dates, there are no published reports on its continuous path planning. In this thesis, a new type of feature, the curved reflection axis, is formulized for free-form surfaces to provide a powerful algebraic model for planning sweep scanning path. It is defined to be a smooth 3D curve, any point of which is the local reflection center of some surface points. The simulation results convincingly prove that the scanning path computed according to a curved reflection axis can maximize the measuring region of one continuous sweep scanning to provide a much faster surface inspection.

Part-in-whole matching, which is to recognize query shapes as sub-parts of a target intact 3D model, plays a pivotal role in a large number of engineering applications. Most state-of-the-art approaches are not practical because two shapes of different scales and sizes cannot be compared by their local shape descriptors and their interest points usually have low correspondence to human judgment. To solve these two issues, a novel local shape descriptor is proposed based on the concept that the evolution pattern of the geodesic iso-contour’s length is a good representative for surface features, and a new scheme of interest points sampling is designed by detecting points sparsely distributed on salient features, transition border boundaries and tipping points on free-form surface. In our experiments, the proposed part-in-whole matching approach with our local shape descriptor and interest points is found to outperform many existing approaches in matching efficiency and requiring a smaller input region.

Finally, the applications of the developed methods are not limited to sweep scanning and partial matching, but also include all techniques requiring symmetry detection and pattern recognition.