破损陶质文物数字几何虚拟复原方法研究

There would not be any change in quality and no deformation of an artifact which is made by ceramic, even though it is broken, that is, the geometrical shape can be repaired. Aiming at the matching requirement of debris, the project focuses on the space intelligent calculation method of the models, the construction of digital geometric restoration technology system, and making breakthrough in the matching efficiency and quality. Thus the project is of great significance..① In view of the serious erosion, blurred contours and lack of features of damaged artifacts, the project presents the research on a novel method of data-driven feature representation of models, in order to achieve the accurate modeling..② Addressing the problem that the fragmentation set is large, the matching efficiency is low and the matching quality is poor, the project makes the innovation in model matching optimization and calculation system, and implements: (1) strategy optimization: decomposes and moves forward the matching contradiction, introduces the intelligent classification, to reduce the size of candidates; (2) algorithm optimization: study the matching method based on significant features, group intelligence and template guidance, to solve the NP Problem and to improve the matching efficiency; (3) guidance optimization: presents the three-level (measurement, knowledge and relationship) effect evaluation and feedback mechanism, to refine matching iteration and improve the matching quality..③ For the research and public demand, the project supports the service calculation of the model: (1) constructs the missing restoration system, creates the virtual restoration system of terracotta, supporting the digital survival of damaged artifacts. (2) construct the AR / VR platform, provides the virtual-real multidimensional representation of the cultural relics, leads the domain application, highlights the inheritance value of the precious cultural relics.

破损陶质文物破不改质、损不变形、几何形态可修复。针对碎片模型拼接需求，课题研究模型空间智能计算方法，构建数字几何复原技术体系，突破匹配效率与质量瓶颈，意义重大。.①针对破损文物侵蚀严重、边缘模糊、特征受损，研究本体数据驱动的模型特征表示新方法，实现精准建模。.②针对碎片集大且匹配效率低质量差的瓶颈，创新模型匹配优化计算体系，实施⑴策略优化：将拼接矛盾分解前移，研究智能分类降低候选集规模；研究目标导引的模板生成方法，解决穷举试拼盲目性。⑵算法优化：研究基于显著特征、群体智能和模板引导的匹配算法，解决拼接NP难，提高匹配效率。⑶指导优化：度量、知识、关系3层次研究效果评价与反馈机制，匹配迭代精化，提高匹配质量。.③针对研究与公众需求，支持模型服务计算：⑴构建缺失修复体系，建设秦俑虚拟复原系统，支撑破损文物数字化生存。⑵构建AR/VR平台，提供文物虚实多维呈现，引领领域应用，彰显珍贵文物传承价值

针对兵马俑数以万计碎片瓦砾拼接需求，陶质文物虚拟修复NP难的问题，提出构建了文物几何计算和形状匹配的计算体系，研究提出了文物模型预处理、几何特征提取、分类和分割、自动拼接、修复和三维模型重建等关键算法，建立了自底向上和自顶向下两种匹配和修复策略，克服了传统穷举试拼方法计算量大、效率低的不足，提高了文物修复的效率和准确性。研究成果主要包括：.（1）提出文物点云压缩、简化、去噪等预处理方法，减少了文物点云模型后处理的规模，提高效率。.（2）提出受损模型的特征提取表示系列算法，提高后续拼接中形状表达与匹配准确率。.（3）提出基于区域和形状特征、基于低维描述符和无监督学习、基于深度学习和模板引导、基于融合注意力机制与多尺度策略的文物点云分类方法。将拼接矛盾前移，缩小候选集合，解决大碎片集穷举试拼接盲目性。.（4）提出基于自适应邻域匹配的点云配准方法、基于断裂面邻域特征、基于调整轮廓线权重和基于特征区域划分等二十多种模型匹配拼接修复算法，构建破损文物模型几何计算和形状匹配技术体系，优化拼接质量。.（5）提出多种修复方法，打破当前模型修复方法中修复裂缝大且难以实现模型表面细节的局限性,实现破损模型的快速修复。.（6）构建交互式文物真实感虚拟仿真平台，突破传统文物展示的瓶颈。虚拟复原实现对实体复原的指导，为文物活化展示传承提供新方法。.项目期间共发表论文91篇，其中一标39篇，SCI收录31篇，EI收录27篇；培养博士后3人，博士8人，硕士30人；申请发明专利25项，其中授权9项；软著6项；企业、团体标准各2项；两项专利转让企业，降低了虚拟角色发型建模的成本和几何模型网络传输的带宽需求；主办承办联合承办领域高峰学术会议及论坛7次、主题会议报告10次；获省奖2项。构建破损文物数字化复原的新体系，设计文物快速建模方法与设备，支持文物的精准建模、虚拟修复，研究成果已应用于兵马俑虚拟复原工程与智慧博物馆建设，提升破损文物的社会文化价值，理论意义和应用效果显著。