Research On Multi-sensor Data Fusion Algorithm Based On D-S Evidence Theory

With the development of IoT,the use of sensors was more and more extensive in the real world.Compared with the instability and low credibility of the single sensor system,the multi-sensor system could detect the information of the target at most.Multi-sensor data fusion was to detect,associate and estimate the inaccurate information which was collected by sensors.It was reduced the uncertain information and got information with the real environment at most.In practical applications,the complex and changeable environment need multi-sensor technology for data analysis,data fusion and decision making.Therefore,it was very important to master the methods and principle of multi-sensor data fusion.On the basis of D-S evidence theory,this paper introduced its theoretical basis,combination rules,analyzed its existing problems and some typical improvement methods.Finally,aimed at the shortcomings of D-S combination rule in handling conflict evidence,an improved method was proposed,and the effectiveness and applicability of the algorithm were proved by experimental simulation.In this paper the method was divided in to two parts: first,the mutual support and the similarity between the evident were calculated,the modified parameters(discount factor)of each evident was measured.The modified parameter(discount factor)was used to modify the evidence and an accurate result was obtained through the D-S combination rules.Second,it used the fusion result as reference evident to calculate the modified parameters(discount factor)of each evident.Iterated and amalgamate until the last two results were converged.In the experimental simulation,two groups of conflicting evidence were compared.The results showed that the method could correctly identify the target in dealing with conflicting evidence and the probability distribution of recognition was more reasonable and higher.Based on the advantages of iterative correction way,a multi-sensor data fusion method based on improved way was proposed.First,the mutual support between sensors was measured according to the membership function.Then,the mutual support degree was used to calculate the reliability and the average credibility of each sensor,the failure sensor was eliminated and the optimal sensor group was selected.Finally,the corresponding evidence was generated by the support matrix and the reliability weight of the sensor which were updated according to the optimal sensor group.The improved method was used to iteratively fuse the evidence until convergence.The final fusion result was used as the sensor fusion weight and data fusion was carried out.The experimental results showed that the multi-sensor data fusion based on the improved method had better fusion results and the adaptive elimination of the failure sensor enhanced its universality.