Evidential Fusion

Evidential fusion is a rapidly developing field focused on combining information from multiple sources while explicitly modeling and incorporating uncertainty. Current research emphasizes the use of probabilistic frameworks, often employing Normal-Inverse Gamma distributions, to represent the uncertainty associated with individual data sources and to fuse these uncertainties into a unified representation using methods like Dempster-Shafer theory or mixture models. This approach is being applied across diverse domains, including medical image segmentation, stereo matching, and remote sensing, leading to improved accuracy and robustness in prediction, particularly in scenarios with noisy or incomplete data. The resulting trustworthy predictions and uncertainty quantification enhance the reliability and interpretability of models, benefiting both scientific understanding and practical applications.