Target Identity Fusion Method Based on Improved DS Evidence Theory
Yi Wang, Desheng Liu, Yang Yang, Jiatong Liu · 2023
Target data detected by multi-sensors often differ from each other due to different sensor performances, resulting in uncertainty of target identity information. Traditional Dempster-Shafer (DS) evidence theory may achieve counter-intuitive outcomes facing evidence conflicts caused by the uncertainty. Aiming for this, a novel target identity fusion method based on improved DS evidence theory is proposed in this paper. Specifically, the method considers the correlation between evidences and the importance of evidences integrally. First, an improved conflict measurement method based on Pearson correlation coefficient matrix is devised to measure the evidence credibility. Next the information volume of evidences is measured by Deng-entropy to indicate the evidence uncertainty. Then the credibility and uncertainty are weighted and combined to modify the basic probability assignment (BPA) of evidences. Ultimately, the modified evidences are fused by Dempster’s combination rule to acquire the final results. Through three numerical analysis examples, this paper demonstrates the superiority of the proposed method over typical previous approaches, which can effectively process evidence conflicts in multi-sensors information fusion.