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

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.