Abstract

This paper presents a parallel processing network for detecting line segments from moving objects. It is founded on two recent developments from our UCLA research group. The first is a radical approach to dynamic edge detection which is event-driven and based solely on the order of passage over triads of equidistant detectors. Simple operations on traversal order, when a translating edges crosses the triad, yield approximate edge direction and velocity. The second foundation is a novel architecture concept for parallel logical inferences using networks of dynamicaly programmable Boolean logic modules (DPLMs). In this approach, Boolean inference data propagate asynchronously and concurrently through a functionally reconfigurable combinatorial network [Vidal83]. Propagation is wavelike and self-timed, proceeding at the maximum speed allowed by component delays. DPLM networks are made of identical modules with a regular topology, well suited to VLSI.

© 1987 Optical Society of America