数据融合中的残差建模分析与融合算法

Track fusion algorithm with residual bias modeling and compensation

多传感器数据融合系统中,传感器之间存在着难以精确建模的系统误差。即便经过校准,仍然会存在残差。残差的量级与随机观测噪声相当,不同的是,残差是一种随时间慢变的系统误差。目前文献中缺乏有效的残差分析建模手段,从而难以提高融合精度。针对上述问题,建立了残差的数学模型,进而提出了残差补偿航迹融合算法。算法将残差增广至目标状态向量,在状态估计的同时完成残差补偿。仿真结果表明,残差补偿算法极大地提高了目标状态估计的精度,显著改善了机动目标的跟踪性能。最后使用雷达实测数据对算法进行仿真,验证了算法可应用于实际工程系统。

In a multi-sensor data fusion system,it is difficult to accurately model sensor bias.Even after registration,there is residual bias in some degree.Residual bias has the same magnitude as random measurement noise,what is different is that the residual bias is slow time-varying.Most literature on residual bias are relatively simple,lack of effective analysis and modeling tools.As a result,it is difficult to improve the estimation accuracy by data fusion.A mathematical model of residual bias is presented,and a residual compensation track fusion algorithm is proposed.The proposed method transfers the residual bias to target state vectors,then estimates and compensates the residual bias.Simulation results show that the residual compensation algorithm can effectively deal with residual effects,which greatly improves the accuracy of target state estimation,especially for those maneuver targets.Finally,real radar data simulation is carried out and verifies that the residual compensation algorithm is applied to practical engineering environment.