不同精度冗余数据的融合

Fusion of Different Accuracy Redundant Data

针对融合误差的最大值和数学期望,提出了一个评判数据融合方法优劣的标准.随后,提出了一种新的数据融合方法,扩展加权平均法.当待融合数据为两个时,通过理论分析得到了计算融合参数的公式.当有更多的数据参与融合时,通过数值仿真得到了该方法的各个融合参数.该方法能解决最大似然估计法所难以解决的均匀分布数据的融合问题,且具有比包括最大似然估计法在内的其它三种有代表性的数据融合方法更高的精度.

This paper presents criteria to evaluate different approaches of redundant data fusion; the criteria mostly concern the expectation and the maximum of the fusion error. A new fusion approach for multiple data is also presented, as the data have different accuracies. This approach is an extension of the weighted average. Through theoretic analysis, we obtain the formula to calculate the parameters for fusion of two uniform distribution data. Through Monte Carlo method, we get the parameters when we fuse more data or other distribution data. Our extended weighted method can fuse uniform data that cannot be handled by maximum likelihood approach. Its result is compared with other three representative fusion algorithms： maximum likelihood, optimal weighted average, and HILARE method. Comparison shows that our approach is better than all weighted average approaches; it has the smallest expectation and smallest maximum of the fusion error in all the four approaches.