Crowd Density Estimation and Crowd Behaviour Analysis

Abstract

With real time surveillance being the need of the hour, Crowd Density and Crowd Behaviour Analysis can result in better protection and increased quality of services being offered. Urban Planning, crowd estimates, and quick responses to emergencies are some of the applications of this study. Furthermore, this study can be implemented to streamline crowd movements in crowded places and military applications. Many implementations have been presented in the same area, however, different environments might have noise, occlusions and cluttered areas can increase the complexity to analyse crowd density and crowd behaviour accurately. This study proposes a solution, which uses two different models to predict crowd density and crowd behaviour separately. For estimating crowd density, I have implemented MCNN architecture[9], which takes into account the scale variation and has given accurate results for crowd density estimation. Furthermore, another model is used to analyse crowd behaviour, which uses crowd movement, heat maps and energy graphs[10] to precisely estimate the crowd behaviour.