基于高斯滤波与均值聚类的异质多源传感器数据加权融合

Weighted Fusion of Heterogeneous Multi-Sensor Data Based on Gaussian Filtering and Mean Clustering

异质多源传感器之间工作频率存在差异,导致数据之间的一致性较差,加权融合后的观测误差较大,因此提出基于高斯滤波与均值聚类的异质多源传感器数据加权融合方法。采用高斯滤波对异质多源传感器数据空间单元格进行划分,建立基于单元格的最佳连通域,保留传感器内部数据,完成传感器数据的高斯滤波平滑处理。引入均值聚类对异质多源传感器数据进行一致性处理。通过免疫粒子群搜索最优权重和参数,利用最优权重和参数完成异质多源传感器数据加权融合。仿真结果表明,所提方法能够降低融合后传感器数据的观测误差与均方误差,观测误差与均方误差最小值均为0.002。因此,说明所提方法提高了融合后异质多源传感器数据的可利用性。

There are differences in the operating frequencies between heterogeneous multi-source sensors,resulting in poor consistency between data and large observation error after weighted fusion.Therefore,a weighted fusion method based on Gaussian filtering and mean clustering is proposed for heterogeneous multi-source sensor data.Gaussian filtering is used to divide the data space cells of heterogeneous multi-source sensors,establish the best connected region based on cells,retain the internal data of sensors,and complete the Gaussian filtering smoothing of sensor data.Mean clustering is introduced to deal with the data consistency of heterogeneous multi-source sensors.The optimal weights and parameters are searched by immune particle swarm optimization,and the heterogeneous multi-source sensor data weighted fusion is completed by using the optimal weights and parameters.The simulation results show that the method can reduce the observation error and mean square error of the fused sensor data,and the minimum values of the observation error and mean square error are both 0.002.Therefore,the availability of heterogeneous multi-source sensor data after fusion is improved.