Multi-target tracking system based on a joint transform correlator and a neural network algorithm

Abstract

Multi-target tracking systems have been under intensive study and in use for many years for a number of applications. More recently, emphasis has been placed on the application of real-time multi-target tracking procedures to situations involving large number of targets. In this paper, we describe a hybrid optoneural system to tackle real-time multi-target tracking. In our system, an optical joint transform correlator is used for real-time adaptive target tracking. But, because of many correlation signals in multi-target tracking problem, a data association algorithm can be used to associate each of the peak's correlation signals to the correct trajectories of the target's motion. Since the computational load on the conventional tracking algorithms increases rapidly with the number of targets tracked, in this paper, a new simple Hebbian learning algorithm is introduced to obtain the respective moving probabilities of the multi-target in real-time, and the data association is achieved in Hopfield optimization neural networks.