Crowd behaviour analysis and anomaly detection by statistical modelling of flow patterns

In this paper, we investigate the crowd behaviours and localise the anomalies due to individual's abrupt dissipation. The novelty of proposed approach is described in three aspects. First, we create the spatio-temporal flow-blocks of the video sequence allowing the marginalisation of arbitrarily flow field. Second, the observed flow field in each flow-block is treated as 2D distribution of samples and mixtures of Gaussian is used to parameterise the flow field. These mixtures of Gaussian result in the distinct representation of flow field named as flow patterns for each flow-block. Third, conditional random field is employed to classify the flow patterns as normal and abnormal for each flow-block. Experiments are conducted on two challenging benchmark datasets PETS 2009 and UMN, and results show that our method achieves higher recognition rates in detecting specific and overall crowd behaviours. In addition, proposed approach shows dominating performance during the comparative analysis with similar approaches.