Modeling and Recognition of Arbitrary Curved Objects from Image Contours

这是该奖项为期三年的第一年。该研究解决了下一代基于模型的视觉系统，该系统应该展示从大型模型数据库中选择的复杂弯曲物体的单眼图像数据的识别。这需要在形状表示、建模和匹配方面取得重大进展。本文研究了一种新的形状表示方法，并将其应用于距离图像和视频图像的自动模型构建和目标识别，包括面向图的构建、姿态计算和模型数据库的索引。在本研究中，任意光滑的三维形状由一组特征表面曲线表示，这些特征表面曲线具有决定图像轮廓结构及其灾难性变化的特性，并且其中一些曲线与单眼轮廓特征（如屈折和bitangents）之间存在天然对应关系。

This is the first year of a three-year continuing award. The research addresses the next generation of model-based vision systems, which should demonstrate the recognition in monocular image data of complex curved objects selected from a large database of models. This will require major progress in the areas of shape representation, modeling, and matching. This research focuses on a novel shape representation and its application to automatic model construction from sets of range and video images and to object recognition, including aspect graph construction, pose computation, and indexing into a database of models. In this research, arbitrary smooth 3D shapes are represented by a set of characteristic surface curves having the properties that they determine the structure of the image contours and its catastrophic changes, and there is a natural correspondence between some of these curves and monocular contour features such as inflections and bitangents.