Research Of Technologies For Situation Assessment In Data Fusion

Situation assessment as a part of data fusion, its realization needs to establish appropriate inference algorithms and system model. In this thesis, the functions of situation assessment to be realized are divided and the technologies to implement these functions are studied, which all provide, to a certain degree, research foundation for designing and realizing applicable situation assessment system.In the first part of this paper, a three-level situation assessment functional model, which consists of situation perception, situation recognition and situation projection, is presented and it leads us to a functional decomposition of a multilevel situation assessment process. The essential characteristics and reasoning mode of situation assessment are analyzed in detail with some examples present and the common essence of situation assessment and its reasoning frame is concluded. After discussion of algorithms of situation assessment, a universal template matching algorithm is brought forwards. Based on all above, the key technologies to assess situation are studied. First a multi-level and multi-hierarchical blackboard model is put forward and how to design and realize data structure and control mechanism of the model is discussed particularly. Then the theory of plan recognition is studied and the process of situation assessment to be that of plan recognition is also discussed. A ''wait and see" tracking plan recognition algorithm is presented. In addition, the application of Bayesian network to situation assessment is discussed and a kind of static template model can be realized in this way. The disadvantage of this static model is analyzed and the algorithm of constructing of a dynamic Bayesian network model is expected. At last situation assessment software system architecture is presented which is based on expert system and blackboard model.