Numerical analyses of modified DS combination methods based on different distance functions

Abstract

In this paper, we give a systematic study of conflict information fusion methods based on different distance functions. With the rapid development of multi-sensor technology, the multi-sensor fusion has gained widespread attention. On the one hand, due to unreliable sensor accuracy, possibly external influence, and artificial disturbance, multi-sensor system always has inevitable uncertainty. Hence, we adopt an effective uncertainty management measure — DS evidence theory to establish a multi-sensor fusion method. On the other hand, potential sensor failure or error will cause the multi-sensor information becoming conflict, and further affect the fusing precise of multi-sensor fusion system. To address this problem, we firstly introduce 6 different distance functions C Euclidean distance function, Jousselme distance function, Minkowsky distance function, Manhttan distance function, Jffreys distance function and Camberra distance function to correct the conflict information, which can improve the reliability of conflict information. And then, according to the above distance functions, we separately modify the multi-sensor fusion method based on DS evidence theory to build an information fusion method that suitable for conflict information, which can enhance the reasonability and rationality of multi-sensor fusion system. At last, to give a synthesized analyses of these modified conflict information fusion methods based on different distance function, we adopt 3 groups of conflict information in the multi-sensor fusion system. Experimental results and analyses demonstrate that compared with the traditional DS combination method, these modified DS combination methods based on different distance functions correct the conflict information before information fusion processing, which all have better fusing results. Thus, these proposed algorithms in this paper has certain practicability and availability.