A Fuzzy Logic-based Approach to Detection of Abnormal Crowd Behaviour

The paper presents an approach to detection of abnormal crowd behaviour at microscopic level. The characteristic individual and group motion patterns are specified with fuzzy predicates. The human interpretation of an video sequences of an abnormal crowd behaviour, based on common-sense knowledge, is mapped into fuzzy logic functions. Based on the evaluation of these functions, the abnormal crowd behaviour is detected. The usability of the proposed approach is tested on video sequences with ground truth annotations. The preliminary experimental results showed that the approach supports an efficient representation of expert knowledge and detection of anomaly crowd behaviour.