Multi-sensor Synergy Life Detection Technique Based on Extended D-S Rule

Abstract

A kind of information fusion algorithm is designed based on extended D-S rule for multi-sensor synergy detection, in order to solve the multi-source information fusion problem of uncertainty and conflict in life recognition. First, multi-sensor synergy life detection platform structure is given. Second, the D-S rule and extended D-S rule are used to solve decision level data fusion. Finally, we complete the information fusion recognition of infrared information and acoustic. By comparative analysis, the effectiveness of extended D-S rule is verified in the life detection. Introduction The multi-sensor synergy life detection technology is a life information fusion which obtained from multi-source. This technology can produce more effective and more accurate estimates of the living organisms than any single source of information [1]. Similar or different kinds of sensors are used to provide comprehensive information, which can make up for the limitation of single sensor [2][3]. The D-S theory is suitable for information fusion of without a priori, and it has advantages on the uncertainty representation, measurement and combination, at the same time it accords with human reasoning decision-making process. But in the case of high conflict evidence, evidence theory can produce rather counter-intuitive conclusion. A lot of literatures have indicatedd that it is caused by composition rules, and improved composition rules, but the effect is not ideal [4-10]. In this paper, we extend the D S fusion rules, and regard the conflict as a form of information. Research shows that extension rules can not only make high conflict evidence for reasonable fusion results, but also it can merge the conflict information even which the general basic probability is 1. The algorithm has strong robustness, and has no additional conditions compared with D S rules. The simulation results show that the extended D-S rule not only can expansion the use range, but also the algorithm is effective. D-S combination rules and the extended D-S rules Combination rules reflect the combination of a law of evidence. In the following, we assume that Θ= {θ1, ... , θn} is a finite set (called frame) of n exhaustive elements. If Θ= {θ1, ..., θn} is a priori not closed (Θ is said to be an open world/frame). Bel1 and Bel2 are reliability functions of the same recognition framework on Θ, m1 and m2 are respectively corresponding to the basic reliability distribution. If A⊆θ and m(A)>0, then A is called focal element. The focal elements are expressed as A1, ..., Ak and B1, ..., Bl. If ( ) ( ) 1 2 1 < ∑ Φ = ∩ j i B A j i B m A m , then : ( ) ( ) ( ) ( ) ( ) φ φ