Home // ADVCOMP 2018, The Twelfth International Conference on Advanced Engineering Computing and Applications in Sciences // View article

An Innovative Crowd Segmentation Approach based on Social Force Modelling

Authors:
Yu Hao
Zhijie Xu
Ying Liu
Jing Wang
Jiulun Fan

Keywords: crowd segmentation, motion prediction, social force model

Abstract:
The effective segmentation and separation of groups in a crowd serves a crucial role in the video analysis for crowd behavior understanding. Conventional crowd segmentation techniques rely on the spatial distribution or temporal trajectories of individual agent in the crowd. However, psychological influences among subjects in a high density crowd are often ignored. In this research, an enhanced Social Force Model (SFM) is developed to define in-crowd motions among agents and its affiliated groups. An innovative SFM descriptor - the Grouping Center - is modeled as a key feature for aiding accurate crowd segmentation. During the process, extracted optical flows are mapped to the detected agents, then the SFM components such as the desired force and the repulsive force are calculated. Experiments show accurate segmentation can be achieved even when agents from different groups are heavily (spatially) mixed up. Random crowd scenarios generated by a commercial game engine are used for crowd behavior analysis and model evaluation. Evaluations indicate that the new SFM model and its related-techniques are capable of handling complex crowd scenarios with latent semantic meanings.

Pages: 18 to 23

Copyright: Copyright (c) IARIA, 2018

Publication date: November 18, 2018

Published in: conference

ISSN: 2308-4499

ISBN: 978-1-61208-677-4

Location: Athens, Greece

Dates: from November 18, 2018 to November 22, 2018