论文部分内容阅读
在典型的多目标跟踪中,随机干扰的出现给量测源带来了不确定性。因此需要用数据互连技术使每个量测与相应目标互连或当出现杂波或虚警时放弃该量测。本文介绍使用Hopfield网络的基于神经网络的多目标跟踪算法。通过把数据互连问题的约束和著名的“旅行推销员问题”(TSP)的约束相比校,导出Hopfield网络的能量函数。通过模拟退火过程使能量函数最小化,由此计算数据互连概率,并应用于每个目标的卡尔曼滤波器跟踪器。将所提出算法的性能与传统技术相比较。仿真结果表明:所提出的神经网络跟踪器与联合概率数据互连滤波器相比较有令人满意的性能。
In a typical multi-target tracking, the presence of random interference brings uncertainty to the measurement source. It is therefore necessary to interconnect each measurement with a corresponding target using data interconnect technology or to relinquish the measurement when clutter or false alarms occur. This paper introduces a neural network based multi-target tracking algorithm using Hopfield network. The Hopfield network’s energy function is derived by comparing the constraints of the data interconnect problem with the well-known “Traveling Salesman Problem” (TSP) constraints. The energy function is minimized by simulating the annealing process, whereby the data interconnect probabilities are calculated and applied to each target Kalman filter tracker. The performance of the proposed algorithm is compared with the traditional one. The simulation results show that the proposed neural network tracker has satisfactory performance compared with the joint probability data interconnection filter.