Situation Graph Trees

In many domains high-level descriptions to represent the status of an environment is desirable. Describing a situation requires to conceptualize the knowledge about the possible actions of the actors involved in the environment and their possible interactions. Conceptual descriptions of different application domains like traffic analysis[1], parking lot security [2,3], human behaviour recognition [4] The conceptualization can proceed from simple descriptions (simple events) to complex descriptions (complex events). Following relations, concepts can be aggregated into more complex concepts. Hence, an event can be described as a sequence of simple events. To allow such an incremental description of the events two main process are required: a) Modelling of the behaviours and b) the reasoning engine. Different approaches like Bayesian networks [5], Hidde Markov models [6], SVM [7] have been recently introduced. However, in complex environments such a models are difficult to apply as they would require a large amount of data for training the models. In such cases, a formalism to model behaviours by means of temporal and semantic relationships or specialization of fundamental concepts can be useful. Situational Graph Trees (SGTs) represent such a modelling tool.