Classification, Stitching And Parallel Algorithm Of Relics Fragment

A Cultural relic is one of important objects for research of archaeology because archaeologist can obtain information such as production, life, science &technology, art and so on from cultural relics. Because of nature and human-generated forces, there exist hundred and thousands of fragments at archaeological excavation sites. How to quickly reconstruct complete cultural relics from fractions is a question, which archaeologists are interested in. At present, this work is finished mainly by hand. It is not only time-consuming, but also damaged to cultural relics. So it is emergent to develop a computer-aided reconstruction system.Generally, computer-aided classification and stitch are primarily performed on the basis of geometry, texture etc information of relic fragments. However, as relic fragments often exist in large quantities; ordinary personal computer is difficult to complete the work in a short period of time. So parallel algorithm is necessary to be adopted to improve the procedure.Based on computational geometry, computer graphics as well as archaeological knowledge, we studied and designed a classification system of relic fragment of rotation body. First, 3D scanner Inspeck is used to retrieve 3D point cloud information, with which corresponding rotation axis and generatrix of these fragments are estimated. Then we matched, classified and archived these fragment according to the shape similarity of generatrix. Finally, for the fragments in the same document, we stitched them by rotation and translation and presented an improved algorithm, which greatly reduced the original algorithm's time complexity.Main contributes are as follows:1. Complete the parallel classification and stitching system of relics fragment.As it costs plenty of time to piece together a large amount of relics fragment, parallel processing which make use of MPI message passing mechanism is adopted. Principal-subordinate mode is employed in parallel algorithm, which means a node is picked out as principal node and the others are subordinate nodes. The principal node takes charge of distributing relics fragment to subordinate nodes and get information about axis and generatrix from subordinate nodes. Concretely, firstly the principal node classifies relics fragment base on generatrix similarity. Secondly, the principal node assigns congener relics to subordinate node. Thirdly, fragment is stitched in subordinate node. Finally, results are returned to the principal node.2 A new classification algorithm of fragment is presented, based on the similarity of generatrix of relic fragment.According to the feature point and sampling curvature of the fragment generatrix, we compare the similarity of generatrix, and classify those matched within the threshold value set in advance into one document. The algorithm solutes the classification problem that texture features of fragments is in the loss.3. An improved profile computing of fragment and stitching algorithm is presented.According to features of 3D point cloud surface periphery, we give an approach to compute the contour of relic fragments. Based on the rotation features of relics, a new stitching algorithm is given, which greatly improved efficiency of the original.