Learning Spatio-Temporal Patterns for Predicting Object Behaviour
Neil Sumpter, Andrew J. Bulpitt · 1998
Rule-based systems employed to model complex object behaviours, do not necessarily provide a realistic portrayal of true behaviour. To capture the real characteristics in a specific environment, a better model may be learnt from observation. This paper presents a novel approach to learning longterm spatio-temporal patterns of objects in image sequences, using a neural network paradigm to predict future behaviour. The results demonstrate the application of our approach to the problem of predicting animal behaviour in response to a predator. 1 Introduction The recognition of spatio-temporal patterns within a scene is an important facet of computer vision research. Future behaviour of an object, in terms of its motion and appearance, can be implied through a learned model of previous behaviour. Short-term predicitions of likely object motion and deformation over one time-step allow objects to be tracked robustly through a scene. This has been achieved successfully using a Kalman...