Research In The Fine Grain Modules Extraction And Management Of The Information Fusion Filter Algorithm

This thesis is based on the multi-sensor information fusion theory and the information fusion algorithm management theory. According to the condition that the fusion result can’t be improved by only adjusting the algorithm, we focus on disassembling the algorithm in fine grain modules or parameters, and by reselecting or recombination the modules or the parameters on the feedback of the fusion result, we complete the optimization of the whole fusion algorithm. Especially when lack of a lot of prior knowledge of the objects or no existing algorithm is fit for the very situation, we can use the method of algorithm fine grain module or parameter recombination to replace the existing feature modules in the algorithm to complete the algorithm optimization and adjusting.According to the soft and hard decision management and the number of the model in the algorithm, we divide the method into two directions:1) Hard decision-single module:In the single module algorithm we focus on the relationship between ψ parameter in the filter algorithm and the innovation in the filter process, and then deduce the sensitive metric A(k+1) by the known parameters to feedback the fusion result to finish the recombination and optimization of the single module algorithm.2) Soft decision-multi-module:In the multi-module algorithm, we introduce the word "module group", and make the module group self-adaption by description of the module matching attribution of the module probability, so that we complete the recombination of the module group.