Neural network approach in multiple target tracking

Abstract

Solutions of the multiple-target tracking (MTT) problem using programmatic as well as nonprogrammatic approaches are discussed. Special attention is focused on the application of neural networks from both analysis and implementation points of view. Results indicate that Hopfield's traveling salesman problem paradigm is applicable to the MTT problem. The neural network approach has the advantage that a large number of targets only requires a larger array of neurons as opposed to the programmatic approach which requires an exponential increase of computer resources. In the near future neural networks may be implemented using conventional IC technology. In the not-too-far distant future, implementations will be using new technologies such as GaAs, VLSI, and VHSIC (very high speed integrated circuit).< >