A Graphical Simulator for Modeling Complex Crowd Behaviors

Abstract

Abnormal crowd behaviors within varied real-world settings could represent or pose serious ongoing situations public safety. However the video data needed for relevant analysis and research are often difficult to acquire due to security, privacy and data protection issues. Without large amount of realistic crowd data, it is difficult to develop and verify crowd behavioral models, corresponding event detection algorithms, and never mention the necessary test and evaluation work. This paper presented a synthetic method for generating crowd movement and dynamic data based on existing social and behavioral studies, graph and tree search algorithms and, game engine techniques. The two main outcomes of this research: 1) the categorization of entity-based Crowd Behavior Types based on a linear aggregation model; and 2) an innovative agent behavior model based on A-Star (A*) path-finding algorithm and an enhanced Social Force Model. A Spatial-Temporal Texture (STT) technique has been adopted for the evaluation of the model's effectiveness. The experimental results have shown the visual and behavioral pattern similarities of the simulated crowd scenes against their real-world recordings.