Dissimiliar-sensor Data Fusion Based On Large Amount Of Information

Due to the advent of the sensor technology and communication technique, multi-sensor systems have recently attracted considerable attention, especially, those with diversified complex application background. The emerging interest of research into multi-sensor information theory is viewed as timely since the multiform, massive, complex and real-time processing of information in multi-sensor system has gone beyond the capacity of traditional information processing. State fusion estimation is an important study field in the information fusion theory, it is usually applied in exact estimating systems.This paper considers state fusion estimation of multi-sensor information fusion theory, mainly dealing with how to estimate the system state exactly by multi-sensor, we have done in-depth studies on sensor management and data fusion algorithm. The main contributions are:A sensor management method based on Quantum-behaved Particle Swarm Optimization (QDPSO) is presented, through analyzing the theories of multi-sensor management. The objective function of sensor management synthetizes the information gain and the process time. Then the formula for multi-sensor data fusion and the probability distribution derived using the maximum entropy algorithm and Bayes estimation. Then, a data fusion algorithm when all the sensors measure the same parameter more than once based on the Recursive Least Squares (RLS) criterion is present. The relationship among the weights of respective sensor and the variance of the measured error has been deduced, and a learning algorithm for variance estimation has been proposed as well. Finally, through analyzing the measurement model of multi-sensor, the weighted fusion formula of parameters fusion estimation from multiple dissimiliar-sensors is surveyed on the principle of least squares, and the scaling coefficients of various sensors weighted fusion has been obtained.