Implementation of Data Fusion using Dempster-Shafer Evidence Theory

In this paper, Sun Quan proposes a new combination formula based on Dempster-Shafer evidence theory to achieve data fusion. Data fusion is a technique that integrates information from multiple sources to enhance data quality and reliability. The method employs Dempster-Shafer evidence theory - a probability-based reasoning approach particularly effective for handling uncertainty and incomplete information. From an implementation perspective, the method combines evidence from different data sources using a novel combination rule that calculates basic probability assignments (BPA) and applies Dempster's rule of combination. The algorithm typically involves these key steps: 1) defining frame of discernment, 2) establishing mass functions for each evidence source, 3) computing conflict factors between evidence sets, and 4) applying the combination rule to merge evidence while managing conflicts. Specifically, the methodology synthesizes evidence from disparate data sources to obtain more accurate and robust results. The approach finds applications across multiple domains including computer vision (for multi-sensor image fusion) and natural language processing (for combining outputs from different linguistic models). The research contribution significantly improves the efficiency and precision of data fusion processes, providing new theoretical frameworks and methodological approaches for advancing data fusion research.