面向陶瓷大数据的三维模型构建及检索研究

Ceramic is the combination of art and technology. Ceramic products have an origin history and become an important cultural heritage. And along with the rise of the mobile Internet industry, our life become more and more convenient. Benefiting from the great cultural heritage of ancient ceramic, the study of 3D model retrieval of ceramic products owns significant socio-economic value. This project will classify ceramic feature according to the traditional ceramic designing rules and interrelated data, on this basis: (1) to handle the problem of continuity and freedom in traditional modeling methods , a unified prototype-based sparse ceramic automatic modeling framework is proposed. The framework is based on deep learning. Sparse rules constraints were added on the basis of AutoEncoders, The framework does not depend on specific methods, which can improve generalization capabilities; (2)to compress feature, a unified polymerization descriptor framework based on supervised sparse coding is proposed. Based on the feature constraint model, we will make use of a high-quality model integration technique to integrate ceramic feature. Furthermore, aiming at different needs, we will adopt various feature fusion methods, not only does this simplify the process of ceramic design, but also it ensures the desired properties of integration modeling; (3) we will investigate the 3D digital model built upon ceramic feature information with respect to users' requests, this shall facilitate innovative designing of ceramic products. In this project, a software system for 3D model retrieval of ceramic based on compact descriptor, which will integrate the above theoretical research into practical application scenarios. This project will not only make the contributions to ceramic designing, but also set up the relationship between ceramic enterprises and users to promote the development of ceramic industry.

陶瓷是科学和艺术的结合，我国陶瓷历史悠久、博大精深，而近年来移动互联网产业迅速崛起，给我们的生活带来的很多便利。研究如何利用智能设备进行陶瓷三维模型检索具有重大社会经济价值。本项目将根据陶瓷传统设计规则与数据对陶瓷特征进行分类，在此基础上：(1)面对传统建模方法存在连续性与自由性冲突问题，对陶瓷特征进行建模，实现基于原型的稀疏陶瓷自动建模方法，该框架使用机器学习，不依赖具体某类建模方法，可以提升泛化能力；(2)对特征描述子进行压缩，设计基于有监督编码的聚合描述子统一框架，结合零阶和高阶聚合描述子并克服各自的缺陷。（3）研究根据陶瓷特征信息创建出符合用户需求的陶瓷制品三维数字模型，降低陶瓷制品创新设计的难度。本项目计划研发出陶瓷三维模型检索软件系统，为陶瓷设计服务的同时，也建立起陶瓷企业和用户之间的联系，促进陶瓷产业的发展。

本项目研究陶瓷产品的数字化，根据陶瓷传统设计规则与数据对陶瓷特征进行分类，在此基础上：(1)面对传统建模方法存在连续性与自由性冲突问题，对陶瓷特征进行建模，实现基于原型的稀疏陶瓷自动建模方法，该框架使用机器学习，不依赖具体某类建模方法，可以提升泛化能力；(2)对特征描述子进行压缩，设计基于有监督编码的聚合描述子统一框架，结合零阶和高阶聚合描述子并克服各自的缺陷。(3)研究根据陶瓷特征信息创建出符合用户需求的陶瓷制品三维数字模型，降低陶瓷制品创新设计的难度。经项目组成员通力合作，共计发表学术论文8篇，出版学术专著1部，授权软件著作权1项，项目成果提高了陶瓷产品设计效率与质量，部分成果已在陶瓷企业应用。