Surface acquisition methods are becoming popular for many practical applications inmanufacturing, art, and design. With the growing amount of geometric data, efficient toolsfor matching and recognition of complex surfaces become more important. In order toachieve such efficiency, many existing methods operate on a limited subset of feature pointssampled from the surfaces, often randomly.In this thesis, we introduce an alternative way to achieve the efficiency by detecting a setof salient feature points from complex 3D geometry data. The method builds a scale-spacerepresentation for the input surface and use local extrema of the difference along normaldirection between neighbor scales as salient points (or features). For every feature detected,we define a point signature vector that reflects the variation of local surface normals. Salientpoints and their signatures are invariant to rigid transformation and are stable under surfacevariation. This provides a good basis for a single feature to find its correct match with goodprobability in a large database of features.We show the effectiveness of selected features and their signatures by applying them tosolve several 3D computer vision problems. We first use the features for pairwise surfaceregistration that matches two partial surface scans or matches a partial scan to its CADmodel. The result of pairwise surface matching is used to align multi-view scans of thesame object to reconstruct the complete model. We also use the selected features andtheir signatures for 3D object recognition, and evaluate their performance on both syntheticand real world 3D data with clustering and occlusion. Experiments demonstrate that theproposed features and signatures are robust for the applications.
【 预 览 】
附件列表
Files
Size
Format
View
Feature Points on Point-based Surface and Their Applications.
PDF
download
878987797
028-85220240
