An Improved Multi-sensor Data Adaptive Fusion Method

One of the major problems of multi-sensor information fusion is that sensors frequently produce spurious observations, which have a great impact on the fusion accuracy and are difficult to be modeled and predicted. Bayesian information fusion technology based on entropy theory is an effective method to solve this problem. However, this method needs the integral operation in infinite interval, and the problem of numerical instability is prone to occur. To solve this problem, this paper proposes an improved multi-sensor data adaptive fusion method. In the framework of Bayesian theory, we use the difference between the measured values of sensors to adaptively establish the posterior probability distribution model of the sensor. Combined with the theory of mutual information, the pseudo-measured values can be identified and eliminated in real time, without integral calculation. The simulation and measured data test results show that the proposed method achieves the same results as the simple Bayesian fusion method without any spurious measurement, and the information fusion performance is obviously better than the simple Bayesian fusion method.