Omnidirectional Vision for Recognition and Control

该奖项以探索性研究小额赠款的形式，资助了一项研究，研究全方位视觉传感器的潜在能力，为移动机器人提供指导和控制信息。这项工作将探讨几个基本问题：(1)导航和目标识别的场景解释是否可以在“原始”全向图像上有效地进行，或者是否需要补偿这些图像中固有的扭曲；(2)如何最好地利用全向视觉中的图像不变量来最小化机器人导航中的计算量。探索性误差研究和全向方法的实验验证也将进行。本研究最终可望为移动机器人的全方位视觉控制提供理论基础。

This award, in the Small Grants for Exploratory Research mode, funds an investigation into the potential capabilities of an omnidirectional vision sensor to provide guidance and control information for a mobile robot. The work will explore several basic questions: (1) can scene interpretation for navigation and object recognition be carried out efficiently on "raw" omnidirectional images, or is it necessary to compensate for the distortions inherent in such images; and (2) how best to use image invariants in omnidirectional vision to minimize computation in robotic navigation. Exploratory error studies and experimental validation of the omnidirectional approach will also be carried out. Ultimately this research is expected to develop a theoretical foundation for omnidirectional vision control of a mobile robot.