Survey of Holistic Crowd Analysis Models
Muhammad Taimoor Khan, Armughan Ali, Mehr Yahya Durrani, Imran Ahmad Siddiqui · 2015
The behavior analysis techniques used in computer vision, are mostly targeting individual's behavior. The study on crowd analysis is more focused on counting individuals and devising a management plan for load balancing for vehicles and pedestrians. The recent advancements in vision based techniques have allowed the study of collective behavior of the crowd for valuable information. It targets the application areas where crowds are dense making it impossible to segment individuals separately due to severe occlusion. Since object detection and identification techniques only works in low density crowds, therefore, holistic crowd analysis techniques are considered for this study. Holistic approach does not attempt to separate out the crowd and rather consider it as a single entity. It studies the behavior of the crowd, instead of the individual's behavior. Recently available techniques for holistic crowd analysis are considered for this study. The working of these techniques is described and a comparative analysis is given to highlight their strengths and weaknesses.