Fusion Units Based Evolution Architectures For Data Fusion System

The nervous system of human being is probably the most flexible paradigm of data fusion architectures in multisensors fusion environment. As we presented in the paper, with the evolution procedure accomplished of neurons from the single-celled organisms and multicellular animals that lack nervous system, the evolution course of nervous system could be summarized briefly as: Chaos phase, Equally nerve net, Centrilized network phase and Bilateral symmetry, hierarchical organization and cephalization phase. With the emergence and evolution of network system which is integrated form kinds of networks to realize a certain mission, the corresponding data fusion techniques and architectures also need to be evolving. By comparing the contemporary data fusion model and architectures with the evolution process of nervous system, an evolution architecture of data fusion system was proposed. An intelligent interface with the ability of running internal model is used to finish the tasks of network communication and resource management, therefore, the kernel of data fusion unit can hammer away at the more important tasks, such as data mining, model configuration and fusion strategies deciding, furthermore, with the importing of data marts, some crucial functions of data fusion system, such as learning, evaluation and storages is also enhanced. The evolution architectures for data fusion is quite flexible, in here, the data fusion event turns into selecting an optimal internal model running on the network, and this model will decide the formed dynamic architectures of the data fusion system. Firstly, the communication strategies and resource management strategies based on the evolution architectures are analysized in the paper, then, the multitarget acquisition and tracking problem based on a priori knowledge and sequence images processing technique are discussed in detail. The simulation results show that, the evolution architectures model win the advantage of integrating the research production such as natural language processing, fuzzy logic, neural network techniques, artificial intelligent and intelligent deciding methods easily. In other words, it will take full advantage of the operators' subjective value, provide an intelligent assistant deciding environment to all levels managers, and boost the information processing ability and command & deciding levels. Those advantages are attributable to that the evolution data fusion architectures are founded on the physical base of anthropic thinking and deciding, i.e., the evolution theory of nervous system.