Detecting abnormal crowd behaviors based on the div-curl characteristics of flow fields

Abstract

This study proposes a divergence-curl-driven framework for the perception of crowd motion states. In this framework, the characteristics of a flow field, divergence and curl, are used to analyze crowd states. As a collective motion, the movement of a pedestrian crowd shows coherent structural properties. By using the methods of fluid mechanics and the feature visualization of flow fields, a physical characteristic descriptor of crowd motion is established that can model the motion state in a crowd flow field. Given the significance of the temporal comparison of motion states for detecting changes in crowds, a method based on the temporal context of motion is presented to measure changes in the distribution of the physical characteristic descriptors of crowd motion. This method can be used to calculate differences in the distribution of the flow field’s physical characteristics between each state and measure these subtle continuous changes on the sample points, thereby obtaining a quantified metric of changes in a crowd’s motion state. Experiments on crowd event datasets demonstrate the effectiveness of our proposed framework for detecting crowd state changes and abnormal activity.