Research Of Fragment Fracture Surface Extraction And Matching Key Technique In The Computer-aided Relic Restoration

With the development of information science technology and digitization, cultural heritage protection also entered the digital stage. It is the key aspect of digital cultural heritage protection that computer aided virtual stitching restoration for heritage. Virtual stitching has lots of advantages with very high historical and practical significance including shortening the artificial recovery period, reducing the hand-stitching difficulty of cultural relics, exploring a new information technical way for archeology and providing a safe and reliable solution for cultural relic conservation. This paper takes fragments from the pit K9901of the Terracotta Warriors as research-objects and researches on the virtual stitching technology for non-thin fragments deeply. Mainly includes:(1) A surface segmentation algorithm using minimum spanning tree(MST) based on integral invariants is proposed. This algorithm first calculates volume integral invariants of fragments’ surface vertex under multiple scales, finding sharp and raised vertices as feature points, then use MST and the corresponding pruning strategies to quickly and accurately generate closed surface dividing lines. Next, the fragment surface is divided into several separate sub-surfaces by the depth-first search algorithm. Finally, fracture surface is identified and extracted according to the degree of surface roughness.(2) This paper put forwards a fracture surface matching algorithm based on feature optimization of constraint clusters. At first, the method clusters nearby and similar feature points. Initial matching cluster pairs and initial error are obtained by using greedy algorithm and the convex and concave correspondence of cluster surface after extracting and representing features of constrained clusters. Then by introducing spatial geometric consistency constraints into branch and bound algorithm, the best fragment matching pairs are acquired through the fast and efficiently minimization of distance root mean squared error(dRMS). At last, verify the adjacency relationship between the two fragments by the evaluation of its approximate consistency of spatial transformation at multiple scales. (3) The algorithm, Iterative Closest Points using Invariant Features, which evolved from the traditional ICP algorithm is used to align the fragments. Volume integral invariants integrated into the distance equation of the corresponding ICP point as the weight make itself become a inspired factor which can accelerate the iterative of alignment and reduce the risk of falling into local minimum.(4) Design and implement a prototype system for cultural relic fragments stitching. The system consists fracture surface extraction subsystem and fragment stitching subsystem. The system which is easy to use facilitates the work of compactly stitching for cultural relic fragments.This research is supported by973plan of preliminary study(No.2011CB311802).