基于BP神经网络的ESM/Radar航迹关联  

ESM/Radar Track Association Based on BP Neural Network

作　　者：林洪涛[1] 周焰[2] 程焱[1] 潘恒辉[1] 

机构地区：[1]空军雷达学院研究生管理大队,武汉430019 [2]空军雷达学院预警监视情报系,武汉430019

出　　处：《现代雷达》2009年第4期65-69,共5页Modern Radar

基　　金：国家自然科学基金资助项目(40101019)

摘　　要：在分布式数据融合的框架下,研究了基于BP神经网络的ESM/Radar航迹关联问题。针对单个融合周期中因传感器采样率差异较大而导致的训练后网络泛化能力较差的问题,对航迹曲线进行拟合后重采样,增加训练样本数据量。利用欧氏距离,对关联概率进行二次估计。将神经网络得到的关联概率与二次估计关联概率进行加权求和,确定最终的关联概率。仿真结果表明,改进后的算法能够对航迹的关联问题作出准确的判决。Based on distributed data fusion, the problem of BP neural network based ESM/Radar track association is studied. As a result of greater sampling period difference between sensors in one fusion period, the generalization ability of the trained network becomes poor. The problem is solved by sampling again after the track curves are fit and the number of sample for training is in- creased. The association probability is reevaluated by calculating the Euclidean distance. The final association probability is ob- tained by calculating the weighted sum of the association probability from neural network and the reevaluated association probability. Simulations show that the modified algorithm can make a correct judgement for association problem.

关 键 词：分布式数据融合 BP神经网络 ESM 航迹关联 

分 类 号：TN959.1[电子电信—信号与信息处理]