Recovery Of Motion Parameters Using Optical Flow

Abstract

In this paper, a novel method of recovering 3D structure and motion parameters using only the apparent motion in the 2D image plane (Optical Flow) is applied to an extended sequence of (real) images. The details of the method are briefly reviewed, and a controlled laboratory image sequence used to establish its performance in practice. The magnitude and direction of the camera's linear and angular velocities are determined, together with an estimate of the depth to any point in the image. These are compared with the known (measured) values for these quantities. The effects of noisy data are examined, and it is concluded that the method is robust in the presence of image noise. Results from a longer "real" sequence are also presented-specifically of a camera flying over realistic terrain. The method is stable over time and is relatively efficient. Possible applications of this method to passive navigation and the guidance of a robot arm are also considered.