부산대학교 도서관

A crucial problem of data fusion in distributed sensor networks mainly results from the unknown correlation in the fusion process. With unknown correlation, the common process noises may cause double-counting errors in fusion results. To optimize the fusion results, the common process noises should be eliminated in fusion results. In this paper, the correlation from common process noises are modeled as the mutual part of two estimates. The size of the mutual part is qualified with information entropy, and then the value range of the mutual part is also analyzed with the joint entropy. Finally, the mutual part is estimated to optimize the fusion results and eliminated with covariance intersection. The simulation of target position verifies the improvement of the proposed fusion method compared to the current alternatives.