ITR/IM:3D Shape-Based Retrieval and Its Applications

本研究将探讨3D模型的自动检索与分析方法。它将开发3D形状的计算表示，可以为其建立索引，可以高效地回答相似查询，并且可以稳健地计算感兴趣的特征。接下来，它将构建用户界面，允许未经培训的用户指定基于形状的查询。这将包括使用文本、3D模型、2D草图以及基于约束和规则的高级方法指定的查询。它将结合计算机图形学、计算机视觉和计算几何的元素。基于形状的查询方法的应用将包括互联网搜索引擎、计算机辅助设计、分子生物学、医学和安全。在每个应用中，研究人员将与领域专家合作，了解3D数据库的关键元素，以及需要哪些新方法来进行具有挑战性的形状查询。例如，与分子生物学家合作将有助于开发蛋白质数据库的查询工具，以找到与给定形状匹配的大分子。这些方法将有助于对只有低分辨率电子密度图可用的蛋白质进行分类，并有助于搜索与特定结合位点匹配的蛋白质。

This research will investigate methods for automatic retrieval and analysisof 3D models. It will develop computational representations of 3D shapefor which indices can be built, similarity queries can be answeredefficiently, and interesting features can be computed robustly. Next, it willbuild user interfaces which permit untrained users to specify shape-basedqueries. This will include queries specified with text, 3D models, 2Dsketching, and high-level methods based on constraints and rules. It willcombine elements of computer graphics, computer vision, and computationalgeometry.Applications of shape-based query methods will include Internet search engines,computer-aided design, molecular biology, medicine, and security. In eachapplication the researchers will work with domain experts to understand thecritical elements of the 3D databases and the challenging shape queries forwhich new methods are required. For example, working with molecular biologistswill help develop query tools for the Protein Data Bank to find macromoleculesmatching a given shape. These methods will aid classification of proteins forwhich only low-resolution electron density maps are available, and aid searchesfor proteins matching a specific binding site.