Crowd object detection and classification based on a chaotic dynamic model
Pengwei Liu · Journal of Shandong University · 2010
Multi-object detection and tracking is an arduous task for traditional methods,and objects in crowds are more difficult to deal with.In this paper,segmentation and detection to video crowd flow was realized by a chaotic dynamics based method.The crowd moving system was treated as a chaotic dynamic system,and object in crowds were treated as particles,whose flow is tracked by flow map.As a result,the finite time Lyapunov exponent(FTLE)field was obtained.After morphological processing,region extraction and classification can be achieved.Experimental results indicate the effectiveness of the proposed approach.