For Full-Text PDF, please login, if you are a member of IEICE,|
or go to Pay Per View on menu list, if you are a nonmember of IEICE.
Measuring Collectiveness in Crowded Scenes via Link Prediction
Jun JIANG Di WU Qizhi TENG Xiaohai HE Mingliang GAO
IEICE TRANSACTIONS on Information and Systems
Publication Date: 2015/08/01
Online ISSN: 1745-1361
Type of Manuscript: LETTER
Category: Image Recognition, Computer Vision
collectiveness, link prediction, random walk,
Full Text: PDF(1.3MB)>>
Collective motion stems from the coordinated behaviors among individuals of crowds, and has attracted growing interest from the physics and computer vision communities. Collectiveness is a metric of the degree to which the state of crowd motion is ordered or synchronized. In this letter, we present a scheme to measure collectiveness via link prediction. Toward this aim, we propose a similarity index called superposed random walk with restarts (SRWR) and construct a novel collectiveness descriptor using the SRWR index and the Laplacian spectrum of a network. Experiments show that our approach gives promising results in real-world crowd scenes, and performs better than the state-of-the-art methods.