Qualitative modelling of object behaviour in the dynamic vision system using hidden Markov models

The subject of this paper is cognitive analysis of object behaviour in a dynamic scene. Based on the identified motions of an animal and calculated trajectory attributes we need to analyse, evaluate and reconstruct the behaviour pattern implementing artificial intelligence methods. Dynamic scene analysis has traditionally been quantitative and it generates large amounts of temporally evolving data. The system presented in the paper uses the qualitative modelling method, based on the model of space and time and incomplete domain background knowledge, in the process of discovering qualitative behaviour patterns. The Markov model (HMM) has been used as a model of object behaviour, based on qualitative spatial and temporal representation and conceptual grouping method.