RE-Assembly Of Fragmented Objects And Computer Aided Restoration Of Cultural Relics

Re-assembling broken objects is a prominent and difficult problem in the filed Computer vision, image analysis and pattern recognition. It can be applied in many fileds, such as archaeology, paleontology, art restoration, and so on. More than ten papers about Re-assembling have been published in international journals and conferences in recent years, but they have a common drawback: that is more on solving 2D reconstruction of broken objects than 3D and previous works emphasize the shape matching but the re-assembling, them only deal with ideal fragments that has no thickness, in all about 2D and 3D.however,in the thesis we focuses on the re-assembling of generalization of 2D and 3D. Mainly works as following.1. The method is introduced of fragment digital, surface reconstruction by scattered data, triangulation, mesh simplification and extracting mesh outlines. Especially we discuss the problem about drawing of the contour line of 3-D thickness object. Also we give an efficient algorithm for drawing of the contour line of 3-D thickness objects.2. We give a algorithm for representing 3D boundary curves by polygonal approximation Using Genetic Algorithms. Simple data reduction is first applied. Chromosomes are defined by encoding the point sequence into binary strings to represent the curve. Each polygonal approximation is mapped to a unique binary string. The objection function is defined as the mean square errors between the given digitized curve and the approximated polygon. Three genetic operators: namely selection, crossover and mutation, are constructed to solve the problem. Points of 3-D digitized Curve corresponding to genes of a chromosome, equal to 1 s, are demarcation ones. Experiments shown that this approach can get more accurate result of approximation.3. We give an algorithm for matching 2D polygonal arcs and application. Feature selection is a key problem of matching. The optimal feature sets should have the attribute of geometry and topology. We focus on polygonal arcs junction that is two line segments that meet at a single point. We present junctions of polygonal arcs as feature sets; the benefits of using this feature sets include attribute of geometry and structure of topology of polygonal arcs. 2D polygonal arcs are represented by the feature sets. This representation is invariant to translation and rotation transformation. The trivial matching algorithm is proposed and use to matching 2D polygonal arcs. Experiments with different classes polygonal arcs show that the matching algorithm is efficiency and is robust to digitization errors and noise effects. And can perform well when applied to re-assembling.4. We give an algorithm for matching 2D polygonal arcs and application. For realistic application, near real time matching 3D polygonal arcs is required. For representing and matching 3-D polygonal arcs. The polygonal arcs junction is defined.3-D polygonal arcs are represented by Spherical coordinates sets that are obtained by defined local Cartesian coordinates system of each junction. This representation is invariant to translation and rotation transformation. The set is views as feature sets. The benefits of using this feature sets include attribute of geometry and structure of topology of polygon. The 3-D polygonal arcs matching task is reduce into a 1-D numerical string-matching problem so that the matching is easy and the processing time is greatly saved. Experiments with different classes polygonal arcs and real images show that the matching algorithm produces sufficiently reliable and is robust to digitization errors and noise effects.5.For re-assembling of fragments, the method for representing and matching 3-D Curve is presented.. The Curves are represented by splines fitted through sequences of points extracted from contour. In reparametrization with arcs length, the curvature and torsion is invariant to translation and rotation transformation. The curvature and torsion are views as feature sets. The 3-D curve matching task is reduce into a 1-D numerical string-matching...