On Modeling and Analyzing Crowds From Videos

Abstract The automated analysis of crowds and the identification of crowd behaviors are important for predicting dangerous situations during events, for appropriately designing public spaces, and for the real-time management of people flows. This chapter covers models and algorithms for the analysis of crowds captured in videos for facilitating personal mobility, safety and security, and enabling assistive robotics in public spaces. We discuss the main challenges and solutions for the analysis of collective behaviors in public spaces, which include understanding how people interact and their constantly changing interpersonal relations under clutter and frequent visual occlusions.