静态交互方式引导的物体功能模型的建立和研究

There is a strong belief that useful distinctions between object categories are often based on functionalities of the objects, thus functionality analysis and recognition play a crucial role for object understanding. Recently, an apparent trend in computer graphics research has seen growing interests in high-level analysis and process of geometry data, especially functional analysis of shapes. It is a critical time to make significant advances. Based on the definition of functionality as the application of an object in a specific context for the accomplishment of a particular purpose, we try to propose a static interaction guided method for functionality analysis. Given a collection of shapes belonging to the same object category, where each shape is provided within a scene context, our goal is to extract the properties that are relevant for the functionality of the category, which will be the functionality model. With the learned functionality models for various object categories serving as a knowledge base, we are able to forma functional understanding of an individual 3D object without a scene context. Such an understanding allows us to recognize or discriminate objects from a functionality perspective, to assess functional similarities, and to possibly discover multiple functionalities of the same object. Moreover, it can also facilitate applications such as 3D shape modeling and 3D printing. Last but not least, we will also provide a functional dataset which can be used to evaluate any work on functionality analysis in the future.

功能性一直都被作为区分不同种类物体的主要依据，因此功能性的分析和识别对于物体的理解起到至关重要的作用。然而计算机图形学领域在这方面的研究才仍处于起步阶段，当前正是在功能性分析上做出突破性研究的关键时刻。通过将功能性理解为物体所支持的与周围环境的相互作用关系，本项目拟提出一种由静态交互方式引导的功能性分析方法。给定一组具有相同功能的物体，以及这些物体所处的环境，我们将提取出为了实现这一特定的功能，物体所应该具备的一系列几何形状要求，从而建立起对应的功能模型。在这一功能模型的基础上，我们可以对三维形状进行功能性评估，并进一步将其应用在三维模型建模以及三维打印上，为模型的创新设计和实例化制造提供引导作用。同时，我们将构建并发布一个功能性分析数据库，供同行使用，为今后功能性方面研究带来便利。

本项目围绕着静态交互方式引导的物体功能模型这一核心主题，从功能数据集的采集、功能模型的构建和学习、单个物体的功能性评估、 多个物体间的功能相似性度量、功能性引导的三维建模和打印等方面展开了系统性的研究，取得了一系列创新性成果。相关成果发表均发表在图形学领域国际顶级货权威会议及期刊，包括ACM TOG论文7篇，IEEE TVCG论文2篇，CGF论文2篇，并在此基础上，提交了10项发明专利申请，登记了5项软件著作权，超额完成了预期目标。