多目标数据关联的神经网络方法

数据关联是多目标跟踪的关键问题。基于 Hopfield神经网络的 JPDA是解决这一关键问题的有效方法之一 ,但此方法的难点在于优化系数的整定。提出一种改进算法 ,用于解决优化系数在线自适应整定问题。首先重新构造了李雅普诺夫能量函数 ,接着引入变化的优化系数因子 ,并给出了优化系数求解的迭代公式 ;最后对已有和改进的算法进行了仿真研究。结果表明改进的方法和原有的方法相比 ,一方面具有在线整定优化系数的功能 ,另一方面可以获得和原有算法非常接近的估计误差。

Research on the Algorithm of Multi-target Data Association Based-on Neural Network

Data association is the key problem in the multi target tracking.JPDA based on the Hopfield neural network is one of the effective algorithms used to solve such a problem.However,it is very difficult to fix the optimal coefficients.Considering the existing algorithm,a new algorithm used to fix the optimal coefficients is proposed by this paper.Firstly,a new Lyapunov Energy Function is reconstructed.Secondly,variable factors of the optimal coefficients are introduced ,and the iterative formulas corresponding to those coefficients are presented.At last,a simulation is performed both on the existing algorithm and the modified algorithm.Compared to the existing algorithm,the modified algorithm not only can have the ability to fix the optimal coefficients on time but also can have nearly same estimated effects as the existing algorithm does.All these results can be proved by the simulation.