An Approach to Multi-Sensor Decision Fusion Based on the Improved Jousselme Evidence Distance

Abstract

Multi-sensor systems are able to obtain various measurement data, but their accuracy and reliability are difficult to be guaranteed, thus the decision-makings using these data are likely contrary to the facts. In view of this, an approach to multi-sensor decision fusion based on improved Jousselme evidence distance is proposed in the framework of D-S evidence theory. By rationally dividing the similarity Jaccard coefficient matrix, the evidences about conflicted sensor node are described accurately and their weights are reallocated by correction. This facilitates the final decision fusion. Numerical experimental results demonstrate that the proposed decision fusion approach based on the improved Jousselme distance achieves better performance than some existed approaches and largely reduces the uncertainty of the fused decision. To sum up, our approach not only recognizes the evidence about conflicted sensor node rapidly, but also has less risk of decision-makings.