Estimation of optical-flow in real time for navigation robots

This paper describes the estimation of optical-flow for an object in 3-dimensional coordinates using a specialized artificial retina sensor and simple template matching methods in the polar domain with a camera motion constraint. The artificial retina sensor (ARS) converts the input image to the polar domain directly. Thus, one can apply horizontal edge detection and acquire the optical flow from the horizontal contour using template matching methods only applied to the horizontal axis. However, the estimated optical flow is distributed on all the object contours, without mis-matching error. The authors demonstrate that this method, in conjunction with the ARS, functions in real-time and may therefore be applicable to systems such as navigation robots and other high-speed applications