Identifying Behaviors in Crowd Scenes Using Stability Analysis for Dynamical Systems

Abstract

method is proposed for identifying five crowd behaviors (bottlenecks, fountainheads, lanes, arches, and blocking) in visual scenes. In the algorithm, a scene is overlaid by a grid of particles initializing a dynamical system defined by the optical flow. Time integration of the dynamical system provides particle trajectories that represent the motion in the scene; these trajectories are used to locate regions of interest in the scene. Linear approximation of the dynamical system provides behavior classification through the Jacobian matrix; the eigenvalues determine the dynamic stability of points in the flow and each type of stability corresponds to one of the five crowd behaviors. The eigenvalues are only considered in the regions of interest, consistent with the linear approximation and the implicated behaviors. The algorithm is repeated over sequential clips of a video in order to record changes in eigenvalues, which may imply changes in behavior. The method was tested on over 60 crowd and traffic videos.

An approach to detect crowd panic behavior using flow-based feature

“…The motion information could be optical flow [1], force flow [5], and tracklet [14]. Global flow-based feature describes overall motion patterns in scene, dominant motion magnitude and tendency can be obtained from it.…”

“…The first one could be the pattern from a cross road; the second one could be at a pedestrian lane; the third and fifth one could be formed at a panic scene; and the fourth one could be a pattern on a Roundabout. This figure is provided by Solmaz [1] Individual behaviors, on the other hand, usually exist in a relative small part of the scene, which may be surrounded by dominant crowd's behaviors, or exists in a sparse scene with low crowd density. For example, pocket picking in crowd and trespassing.…”

Large-scale surveillance system: detection and tracking of suspicious motion patterns in crowded traffic scenes

“…In [8], a framework is proposed to identify multiple crowd behaviors through stability analysis for dynamical systems, without the need for object tracking. Abnormal behavior detection is also studied in [9] using spatiotemporal cuboids, in [10] by modelling appearance and dynamics based on dynamic textures, and by learning normal behavior in [11].…”

Bio-inspired relevant interaction modelling in cognitive crowd management

“…For instance, in (Mehran et al 2009) a method for crowd behaviour analysis based on social forces and optical flow is proposed. More recently, in Solmaz et al (2012) the authors present an innovative method based on people flow estimation. A new abstract viscous fluid field is proposed in Su et al (2012) for detecting crowd events.…”

“…Secondly, it is unrealistic to track every single person in a high density crowded scene, especially if a single camera is available: the visual information gathered by the sensor is simply not enough to accomplish such a task. This remark has led to consider global approaches to crowd monitoring such as in Moore et al (2011Moore et al ( , 2012. At last, the model employed in simulation combines technical and social aspects following the current trend in literature.…”