Research On Digital Geometry Processing Methods In Archaeology Virtual Reconstruction

As a non-destructive method for cultural relic reassemble,virtual merging technology of relic fragments get an extensive attention by archaeologists in recent years.Relic models reconstructed by reverse engineering,can be used for shape matching after applying digital geometry processing and extracting geometric structure features,and be used for reassembling the dissociated fragments further more.In this research,we take the unearthed fragments of Qin-dynasty terracotta warriors as our experiment data,applied several critical techniques in the virtual merging process pipeline,such as point cloud processing,geometric feature optimization,registration,etc.This procedure aimed to solve the problem of damaged large scale data and high computational complexity,so that this technique can be used to instruct manual recovery.Our research has mainly made the following contributions:(1)We proposed an engineering framework to solve the problem of cultural relic reassemble,and by analyzing the characteristics of complete,partial and complex matching problems during virtual merging procedure,we introduced a solution oriented the characteristics of different problems.Based on the solutions of complete matching problem,applied to the partial matching problem;and proposed the feasible solution to solve the problem of complex matching with the useless geometric characteristics;by the analysis and summary,our research covered the field of comprehensive virtual merging problems.(2)We proposed an optimization method for point cloud model,optimized several critical techniques,such as repairing normal,smoothing point cloud,and surface classification,etc.This procedure aimed to solve the problem of point cloud models which is noise sensitive and has complex geometric structure.By the method of normal repairing and point cloud smoothing,we can effectively improve the quality of point cloud,the accuracy of estimation of normal vector and reduce the matching error of point cloud registration stage.To provide data support for the fracture surface matching,we introduced the classification method of surface is based on the analysis of point cloud density,not only overcome the problem of surface classification method is sensitive to geometric structure,but also improved the recognition rate and accuracy of fracture surface classification of cultural models.(3)We proposed a local optimization of the fracture surface matching algorithm,and aimed to solve the problem of the geometrical characteristics of the discontinuous which reduced the fracture matching similarity,we defined local feature descriptors as the solution of adjacency and split position of fragments.In order to calculate the adjacency relationship between fragments,we introduced 0-1 matrix to describe the distribution of geometric characteristics which have better noise immunity in the space,and convert characteristics matching into combinatorial optimization problem under constraints;we proposed a point cloud registration framework based on the penetration test to calculate the pairwise matching positions,the recombinant of multi-piece is a NP hard problem,in order to solve this problem,we converted the reassemble procedure into optimal stitching matrix problem under constraints by defined the feasible solution space of the feasible matrix and the complete model is obtained after iterations.Method of local feature optimization effectively measured the similarity of spatial curved surface,reduced the influence of missing pieces to split process,and improved the robustness of solving the partial matching problems(4)We proposed a fragment matching method under the constraint of multi features,in order to solve the complex splitting problem,such as,model shape is damaged,geometric feature is not available,based on the characteristics of fracture surface we introduced contour curves and the adjacency constraint descriptor to perfect the solution of partial matching problem.By calculating registration error under contour feature constraint,trimming distance offset under two kinds of constraints,and quantizing the merging result to determine the optimal combination strategy;If the geometric features can't be used,we introduced interactive combination method under adjacency constraints and defined constraint descriptor,by limiting the split position of adjacent pieces in a reasonable range to effectively solve the complex assemble problem.Our research was founded by National Basic Research Program(973 Program),and applied in archaeology site in Qin Terra-Cotta Warriors and K9901 excavation program,received good effect.