D-S Evidence Theory for Data Fusion

D-S evidence theory is a mathematical framework for handling uncertainty and conflicting information, widely applied in data fusion systems. Its core principle involves combining multiple evidence sources to derive more reliable conclusions through probabilistic reasoning.

Traditional D-S theory typically employs Dempster's combination rule for evidence synthesis. However, this rule may produce counterintuitive results when high conflicts exist between evidence sources. To address this limitation, Ye Qing proposed an enhanced method in "Systems Engineering and Electronics Technology" that introduces weighting coefficients and a conflict redistribution mechanism.

The key improvements in this method include: First, weighting coefficients quantify the reliability of each evidence source, reducing the impact of low-quality evidence on fusion results. In implementation, this involves calculating evidence credibility metrics using distance measures or correlation coefficients. Second, a redesigned conflict redistribution strategy employs more reasonable probability allocation to prevent traditional method failures in high-conflict scenarios. Algorithmically, this may involve modifying the combination rule using weighted averaging or proportional conflict redistribution techniques.

This enhanced evidence combination method significantly improves the robustness of data fusion systems, particularly suitable for applications like sensor networks and fault diagnosis that require processing multi-source uncertain information. Its advantage lies in preserving D-S theory's uncertainty handling capabilities while solving combination challenges in high-conflict scenarios through weight adjustment and conflict reallocation mechanisms. From an implementation perspective, developers can incorporate these improvements using recursive combination algorithms with conflict detection thresholds and adaptive weighting functions.