Mining Movement Patterns from Video Data to Inform Multi-agent Based Simulation

摘要

Multi-agent Based Simulation (MABS) is concerned with the utilisation of agent based technology for the purpose of running simulations of real world scenarios. The challenge is in encoding the agents so that they operate as realistically as possible. The work described in this paper is directed at the mining of movement information from video data which can then be used to encode the operation of agents operating within a MABS framework. More specifically mechanisms are described to firstly mine "movement patterns" from videos of rats contained within in closed environment and secondly to utilise this information in the context of a simple MABS to support the study of animal behaviour.