Research On Method Of Computer Aided 3D Models Of Terra-Cotta Warriors Fragments Classification

With the development of the computer digitization and visualization technology, cultural heritage protection and inheritance have entered the information stage. Terra-Cotta Warriors fragments, which excavated in the pits of the tomb of the first emperor of Qin, have a large quantity, have incomplete information on the surface, and have complex geometrical shape and structure. It makes the computer aided stitching method low efficiency for original fragments models. In order to effectively solve the high complexity and inaccurate problem of virtual stitching algorithm, we firstly classify the fragments based on body parts, then stitch the fragments belong to the same body part, and lastly, finish regrouping among body parts. At the same time, it can effectively avoid the secondary damage during manual identification classification. Aiming to solve the above problem in fragments classification, this thesis uses the multi-scale characteristic of integral invariants to extract features of fragments models, and use the fast reject schema matching method combined with prior knowledge. This thesis conducts following research works on computer aided fragments classification:1. Propose a feature extraction method based on multi-fielded description of integral invariants. In the thesis, we give the definition and numeric calculation methods of integral invariants, and use the multi-fielded description to express the integral invariants in different scales of given point’s local neighborhood. This description makes better use of the multi-scale characteristic of integral invariants, describes local neighborhood features in different scales and has good robustness.2. Use a regional feature representing method based on multi-layer description to gain the robust representation of regional features of models surface and describe local geometric features. We give the definition of multi-layer description, and then integrate the above multi-fielded descriptions which are ordered by non-decreasing radii to calculate the multi-layer description. This description can effectively reduce the complexity of matching.3. Propose a local matching method based on prior knowledge and the fast reject schema to solve the high complexity and low efficiency problem while matching Terra-cotta Warriors fragments with templates library. Match the given fragment’s features with the model features, note that the model features are dynamically calculated during matching according to calculation requirements. What’s more, compile statistics of matching results, and combine the prior knowledge to identify the classification of the given fragment semi-automatically. The experimental results show that the method we propose has relatively higher accuracy.4. Design and develop Terra-cotta Warriors fragments 3D models classification system. This system is easy to use, with good interaction, and it can identify the classification quickly. It provides effective reference information for computer aided Terra-cotta Warriors virtual recovering work and improves the accuracy of stitching Terra-cotta Warriors fragments.