Panoralllic Range Estill1ation Using a Lllov'ing Eainera and a Speeially Shaped Reflective Surface

1 Introduction A technique is presented for obtaining an omni-directional range map using a panoramic,ffiov-ing visual sensor. The sensor consists of a single , standard video camera viewing a specially shaped reflective surface [ehahl and Srinivasan, 1997b]. The surface has the property that a given change in the angular elevation of view in the external environment maps to a constant radial displacement in the camera's image. Estimation of range is based on the fact that the local, motion-induced deformation of the panoramic image is range-dependent, irrespective of the nature of the deformation. For a translatory motion, the nature of the deformation of the image on a sphere depends upon the the azimuth and elevation of the viewing direction: it is an expansion in the direction of motion , a contraction in the opposite direction, a translation for viewing directions perpendicular to the direction of motion, and a complex deformation along other viewing directions. The range in each direction of view is estimated by comparing the magnitude of the image deformation measured in that direction, with the magnitude of the deformation that would be produced by a reference environment consisting of a "virtual" spherical shell of predetermined radius surrounding the sensor. The local image deformation is measured using a version of the image interpolation algorithm, developed previously in our laboratory [Srinivasan, 1994b]. The technique is capable of delivering omnidi-rectional range maps in quasi-real time. It has potential applications in surveillance systems, as well as in visual guidance and control of autonomous flying vehicles. 246 The development of strategies for navigation of autonomous robots can benefit substantially from the use of wide angle visual sensors. Visual estimation of range has been achieved in a number of ways [Jarvis, 1983], many involving the comparison of images resulting from a known displacement of the sensor in some known direction. Numerous algorithms exist for computing range using two paraxial cameras or optical sensor arrays e.g. This is the so-called 'range from stereo' approach. An equivalent approach would be to use a single camera to capture views from two different locations , and infer range fr~m the motions of corresponding features in the two images. This is the so-called "range from motion" approach, In this paper a technique is developed for the estimation of range in panoramic visual scenes that encompass almost the entire surrounding environment. Panoramic range estimation requires, first of all, a technique for …