基于卡尔曼滤波的动态权值融合

Dynamic weighting fusion based on kalman filter

摘要:

在雷达航迹融合过程中，采用多传感器测量值融合的方法能够摒除单一信息源不全面的缺点。加权平均融合为广泛使用的融合方法，但传统的权值固定的加权平均融合虽然能综合多路传感器信息，却无法自适应的根据测量值优劣倚重更有利的测量信息。因此，本文提出将固定权值改进为动态权值的融合方法，实时改变各路测量信息参与融合的权重。每次融合前，先将多路传感器测量值求简单算术平均后进行卡尔曼滤波，把滤波后的值与各路测量值作差，这相当于对传感器信息的优劣作出预判，每路测量信息的融合权值则与该差绝对值成反比。最后，通过仿真实验证明，该改进方法较之前的加权平均融合明显提高了目标的融合精度。

Abstract:

Multiple sensor measurement fusion can strip away the shortcomings of a single source which the information is not comprehensive in the process of radar track fusion. Weighted average fusion is widely used. The weighted average fusion of traditional and weights fixed can only combine with information from multiple sensors, but not pick out better information adaptively. Therefore, this paper suggests changing the fixed weight to dynamic weight. Before every fusion, calculating simple arithmetic average of multiple sensor measurements, then performing Kalman filter. Making the measurements subtract the values from Kalman filter. That is equivalent to make prediction for distinguishing data of stand or fall. And the dynamic weight is inversely proportional to the value using for prediction. Finally, the simulation experiments prove that the method in this paper can improve the precision of the fusion of target significantly.