Technology Research, Based On Neural Networks For Mobile Multi-target Tracking

With the ability of defending against suddenly of enemy airplanes improving and the density of airplanes increasing, single target tracking can not satisfy the need of modern warfare, the concept of multi-target tracking has developed. Multi-target Tracking is classifying the measurements of numerous sensors according to their sources, and gaining the track of every target, then analyzing the veracity and reliability of the target tracks.This paper applies neural network in target tracking technology through analyzing the function and characteristic of neural network. Firstly, in the view of the drawback of common BP algorithm, this paper studies A Modified BP Network-based Adaptive Tracking of Maneuvering Target.The Joint Probabilistic Data Association algorithm (JPDA) is the accepted effective data association algorithm, but it has high computational load, and it's not a Real-time algorithm. In view of the excellent performance of neural network, especially the ability of parallel disposing, it shows great appliance foreground in the problem of combinatorial optimization. This paper reduces multi-target tracking to be a sort of constraint combinatorial optimization problem to use Hopfield network to get the results, the simulation results show this way can reduce the calculational quantity. Because the Hopfield network is a system of gradient descent, it can get stuck in local minima in searching. It is applied to solve the optimization problem, though it can promise to get a steady balanceable result, it is not the best result of the problem, it is the local best result, it will effect the quality of the result and make the success rate of tracking go down. This paper applies chaotic characteristic to neural network, using the abundant dynamic characteristic of chaotic network to solve the complex optimization problem. So this paper studies A Chaotic Neural Network -based Maneuvering Multi-target Tracking. The simulation results show convergent time of chaotic neural network reduces and the success rata of tracking improves.