IS VISUAL RECONSTRUCTION NECESSARY - OBSTACLE AVOIDANCE WITHOUT PASSIVE RANGING

Is it possible to perform nontrivial visual tasks without using passive ranging? Is it necessary to create, using vision, an accurate 3-D description of the outside world to recognize, navigate, and manipulate? Could a visual system perform obstacle avoidance without reconstructing the depth of the scene in view, but through the recognition of a collision in some amount of time that is quite small and not far into the future? In this article, it is shown that it is possible to avoid obstacles without recovering depth and a detailed algorithm is provided for obstacle avoidance using the spatiotemporal derivatives of the image intensity function as input.