改进CHNN神经网络航迹关联算法  

Improved CHNN Algorithm For Track-to-Track Association

作　　者：柏慧风 高贵明[2] 

机构地区：[1]南京信息工程大学电子与信息工程学院,江苏南京210044 [2]南京船舶雷达研究所,江苏南京210003

出　　处：《现代防御技术》2014年第2期145-149,共5页Modern Defence Technology

摘　　要：航迹关联和航迹融合是分布式数据融合的重点。基于Hopfield神经网络的航迹关联算法是建立在每个传感器都发现相同目标的基础上的,但是在实际应用中往往不是这样。改进了基于连续型神经网络的广义航迹关联算法,进一步地,该算法可以推广到多个传感器的情形。为了加速Hopfield神经网络的收敛速度,重新定义马氏距离。计算机仿真表示,改进算法适用于多传感器航迹关联问题,且适用于密集目标航迹交叉的情形。Track-to-track association and track fusion are the important aspects of distributed data fu- sion. Track-to-track association algorithm based on Hopfield neural network is built on the fact that each sensor detects the same target, but it is not always in the actual situation. An improved algorithm based on continuous state Hopfield neural network （CHNN） is proposed to solve this problem, and this algo- rithm also fits to the situation which has more than two sensors. In order to accelerate the convergence of the Hopfield networks, the Mahalanobis distance is refined . The simulation shows that the improved algo- rithm can successfully solve the track association in complicated environment.

关 键 词：航迹关联 连续型HOPFIELD神经网络 能量函数 马氏距离 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP301.6[自动化与计算机技术—控制科学与工程]