Robot Vision System Based on Log-Polar Image Plane Coordinates

在表达与动态三维透视图像相关的变换时，对数极坐标最近被证明比传统的笛卡尔坐标具有显著的计算优势。像素数量的减少和每像素计算的减少表明了紧凑、高性能系统的可行性，而笛卡尔系统是不可能的。该奖项将提供第一阶段SBIR资金，并将支持机器人视觉试验台的设计，该试验台能够执行有用的视觉功能，如动态深度测绘和相机/车辆运动控制。该系统将结合最近在该领域开发的视觉跟踪和深度映射的新算法，并将这些算法从笛卡尔坐标重新表述为对数极坐标。

In expressing transformations associated with dynamic 3- dimensional perspective imagery, log-polar coordinates have recently been shown to possess significant computational advantages over traditional Cartesian coordinates. Reduced pixel count and reduction in calculations per pixel suggest the feasibility of compact, high-performance systems not possible with Cartesian systems. This award will provide Phase I SBIR funding and will support design of a robot vision testbed capable of performing useful visual functions such as dynamic depth mapping and camera/vehicle motion control. This system will incorporate new algorithms for visual tracking and depth mapping that have recently been developed in this field, and will reformulate those algorithms form Cartesian to log-polar coordinates.***