Multi-sensor Information Fusion Technology Research

As the application of multi-sensor system is more and more important and extensive in modern warfare, the multi-sensor information need to be fusion to get the state estimation, feature data, identification, situation assessment, and threat assessment of the targets. And because of the advantages as farther radar range and good concealment of passive sensor, the research on multiple passive sensor information fusion is paid great attention gradually. This thesis concerns on the technique of data fusion and target recogniton and the design and implementation of the simulation platform of data fusion system. The main contents and results are as follows:Firstly the algorithm of feature data association was researched, the principle and feasibility of association based on feature data was detailedly analyzed. Two effective algorithms of feature data association ware introduced, the'ellipticalthreshold algorithm'based on statistics distance and the'supports threshold algorithm'based on probability and D-S evidence theory. The simulation analysis shows that the performance of'ellipticalthreshold algorithm'is better than the'supports threshold algorithm', and the two algorithms both have good timing performance, and can get high correct rate of association when the features of different targets are not similar.Then the thesis focus on the research of data association algorithms based on azimuth information. The azimuth data association algorithm based on multi-dimension assignment was introduced and detailedly researched, and a otimization algorithm called Lagrangian relaxation technique was furtherly researched to solve the Non-deterministic Polynomial(N-P) hard problem in multi-dimension assignment. By simulation, the Lagrangian relaxation algorithm could get better timing performance under the premise of high correct rate of association. The simulation has also analyzed the factor which can affect the effect of azimuth data association.In aspect of target recognition, technique of feature level identification and decision-making level identification was researched. In identification of feature level, the technique of data fusion and target matching was introduced, and in identification of decision-making level, the fusion technique based on D-S evidence theory was detailedly researched. The simulation result shows indicates that reasonable and credible identification result could be achieved by these methods.Finally, the thesis introduced the design and implementation of the simulation platform of passive sensors data fusion system, and the simulation results based on the simulation platform was analyzed.