Multi-source Data Fusion Approach Based on Improved Evidence Theory

The classical evidence theory can result in paradox in the process of information fusion. To resolve this problem, a multi-source data fusion method based on dissimilarity matrix and evidence theory is proposed. First, using the weighted Euclidean distance, evidence dissimilarity matrix is constructed. Second, dissimilarity between the evidences is measured. Third, using dissimilarity matrix, supporting degree, credibility and weight of evidence are calculated, and the original evidences are modified. Finally, using the improved combination rule, the information fusion is completed. Experimental results show that new method is superior to the existing typical methods in accuracy, discrimination and accuracy of fusion results.