Research On Multi-Sensor Information Fusion And Its Application

With the development of communication technology,electronic technology,computer technology and etc.,the form of information is complex and diverse,the content of information is all-encompassing,and the number of information is unprecedented.In this context,the information obtained from a single sensor is always incomplete,just reflects a certain aspect of the target,and can not grasp the overall content.Therefore,it is necessary to use the multisensor to detect the target from different angles,to obtain more dimensions of information about target,and to fuse the multi-dimensions information,such that it can ensure the integrity and accuracy of the information and help us to make decisions.In this thesis,information fusion technology is studied form two aspects: homogeneous multi-sensor(multiple infrared imaging sensors)information fusion and heterogeneous multi-sensor(Inertial sensors and positioning sensors)information fusion.Firstly,this thesis investigates the multi-sensor information fusion technology,discusses multi-sensor information fusion's basic principle,fusion process,fusion level,architecture and common fusion algorithm.Secondly,on the basis of information fusion theory,this thesis mainly studies the following three aspects.Due to the fact that the classical evidence theory often obtains counter-intuitive result when combing the highly conflict evidences.This thesis puts forward an improved algorithm based on evidence trust factor.First,we calculate the Euclidean distance between the evidence and the entire evidence set.Then,according to the Euclidean distance,we determine the value of the trust factor,and give the corresponding weight of the evidence.Finally,the expected evidence is obtained from the weighted average of all the evidence,and the expected evidence is synthesized.The simulation results show that the proposed algorithm can not only synthesize highly conflict evidences effectively,but also has better performance than classical evidence theory and the other improved algorithms.Aiming at the problem that the basic probability assignment of D-S evidence theory is usually obtained by subjective experience,which leads to the low credibility of decision,this thesis employs the BP neural network to obtain the basic probability assignment.That is because BP neural network has a powerful nonlinear mapping ability,which can reflect the intrinsic relationship of target feature data effectively.The simulation results show that BP neural network not only can obtain the basic probability assignment effectively,but also has the higher recognition accuracy when identifying the target together with the improved D-S evidence theory.Due to the lack of adaptive ability,the conventional Kalman filtering may be divergent in the process of fusing integrated navigation information.For this problem,this thesis puts forward an improved filtering algorithm.which combines the Sage-Husa adaptive filtering algorithm with the conventional Kalman filtering algorithm.and simplifies the processing.In this way,the improved filtering algorithm has the adaptive ability,Moreover,it is more simple than Sage-Husa adaptive filtering algorithm,and the computational complexity is greatly reduced.The simulation results show that the improved filtering algorithm has better convergence and stability compared with the conventional Kalman filter.Additionally,the mean error and the standard error of the navigation information are obviously improved.Thus,it proves the effectiveness of the improved filtering algorithm.