传感器网络中结合分类信息的多目标关联算法  

Classification-aided Multi-target Data Association in the Sensor Networks

作　　者：胡炜薇[1] 秦会斌[1] 胡冀[1] 黄海云[1] 

机构地区：[1]杭州电子科技大学电子信息学院,杭州310018

出　　处：《弹箭与制导学报》2009年第2期203-207,共5页Journal of Projectiles,Rockets,Missiles and Guidance

基　　金：浙江省自然科学基金(Y1080589);教育厅资助科技计划项目(Y20070926)资助

摘　　要：多传感器多目标跟踪系统的关键技术之一是航迹关联,应用于传感器网络时更为复杂。提出结合分类信息的基于图模型航迹关联算法,把目标分类信息和运动学信息相结合来提高关联效果。它利用分类混淆矩阵确定分类结果似然函数,修改基于图模型的航迹关联算法的顶点兼容函数和连线兼容函数。对多传感器网络的多目标跟踪情况仿真,比较仿真结果说明结合分类信息确实提高了关联正确率。Data association is a fundamental and challenging problem in multi-target multi-sensor tracking. It is even more challenging in the context of sensor networks. The paper proposed a classification-aided data association technique based on graphical models to efficiently solve such data association problems. The classification likelihoods based on the confusion matrix is used to modify node compatibility functions and edge compatibility functions of the data association algorithm based on graphical models. The simulations of multi-target tracking in the sensor networks environment, the comparison of results showed classification information improve data association.

关 键 词：航迹关联 跟踪 传感器网络 多目标 图模型 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]