A Real Time Simulated Human Vision System Using Connectionist Models Applied to Target Tracking

associated with partial occlusion and collision is ABSTRACT presented. Fourth. results from the testing arc prcscnted This paper describes an algorilhm and an architecture of a show of the proposcd approach. simulated human vision svstem usine connectionist Finallyv the work to date is summarised. models. An original approach to solve two major problems, occlusion and collision, is proposed for a multiple target tracking system. This approach uses the recall properties of various neural network models. Currently, the system has been built on 64 node Transputer system with 2 M B memory on each Transputer. This system employs both image-intensitybased optical flow methods to compute motions of multiple targets, and token-based methods to track a designated target. Preliminary results using simple multiple objects in an uncluttered scene are also presented. 2. SYSTEM OVERVIEW