A Multi-sensor Target Recognition Information Fusion Approach Based on Improved Evidence Reasoning Rule

Abstract

The Evidence Reasoning (ER) rule extends traditional Dempster-Shafer evidence theory by establishing a new rule to combine multiple pieces of independent evidence with importance and reliability weights. The importance and reliability weight of an evidence source is usually decided by fusion system designers which is subjective. Aiming at solving the evaluation problem of evidence importance and reliability weight in ER rule, a new method is proposed in this paper under the application background of multi-sensor marine target recognition information fusion. The importance weight of evidence source is calculated based on the accuracy of sensor recognition in history observation, while the reliability weight is calculated based on the improved normalized angle distance which measures the conflicting among pieces of evidence. Then the pieces of weighted evidence are combined under ER rule to draw recognition fusion conclusion. The proposed approach improves the ER rule by giving an objective method to measure the importance and reliability weight of evidence. Simulation experiments are conducted, demonstrating that this approach can combine conflicting evidence more effectively. Moreover, compared with other methods, the improved ER rule shares good convergence performance and has higher computational efficiency, which is beneficial for engineering implementation.