(with Shanon X. Ju )
This paper addresses the problem of recovering time-to-contact by actively tracking motion boundaries. First we develop a framework in which the boundaries of objects are automatically detected using the motion parallax caused by the motion of an active camera. We exploit properties of a robust correlation surface to construct a confidence field which expresses the likelihood that a point lies on a motion boundary. Then, the motion discontinuities are interpreted in terms of an extended surface contour. We claim that the discontinuities correspond to surface boundaries in the world, and hence in practice, we can assume such boundaries have spatial coherence and can be modeled with dynamic contours, or snakes. Unlike previous approaches which have relied on manual initialization of the snakes, we develop an automatic initialization scheme based on an attentional mechanism. Assuming an affine transformation of image velocity field, we present a novel method for recovering the affine parameters of the motion boundary via rigid snake tracking. While standard snakes are used to model the spatial coherence of object contours, only affine deformations of the snake are allowed between frames. Finally, time-to-contact of the object is estimated from the affine parameters for deliberately forward motion. We illustrate the behavior of this active approach with experiments on both synthetic and natural image sequences.
Ju, X. and Black, M. J., Time to contact from active tracking of motion boundaries, in Intelligent Robots and Computer Vision XIII: 3D Vision, Product Inspection, and Active Vision, David P. Casasent, Editor, Proc. SPIE 2354, pages 26-37, Nov. 1994, Boston, Massachusetts.