A diffusion mechanism for obstacle detection from size-change information

A new immediacy measure, representing the imminence of collision between an object and a moving observer, is defined. A diffusion process on the image domain, whose initial condition is determined by the motion field normal to the object's boundary, is shown to converge asymptotically to the immediacy measure. Bounds on the convergence time for the diffusion process are described. A network of locally connected neurons, derived from a finite-difference approximation of the diffusion equation, estimates the immediacy measure from normal velocity and boundary information provided by layers of motion measurement and segmentation cells. The network's performance on real image sequences is demonstrated. The proposed scheme is shown to be consistent with observations on the biological visual system.<
>