Counting People in Crowded Environments from Video

Project: Research

View graph of relations


Hong Kong is one of the most densely populated cities in the world. The core urban areas of Hong Kong Island and Kowloon have a population of over 3 million, or a population density of 35,700 per square km. Everyday, large crowds form and flow through public spaces, such as open-air markets, plazas, sidewalks, and subway stations, as the city’s inhabitants go about their daily lives. Understanding where, when, and how large crowds flow in the urban landscape has important applications in resource management (e.g. the allocation of funds and personnel), urban planning (e.g. improving the efficiency of crowd flow), advertising (e.g. targeting ads to larger crowds), security (e.g. the deployment of security personnel), and surveillance (e.g. monitoring maximum occupancy and detecting anomalous events). In this project, we propose to research, develop, and implement a real-time video surveillance system for counting people as they move through a crowded environment. The system will count the total number of people occupying the environment, and will distinguish between crowds moving indifferent directions. The system will also count the number of people crossing a line-of-interest in the video (e.g. entering a specific area in the scene). Finally, using multiple camera views, the system will produce a crowd occupancy map, which details the density of the crowd throughout the environment. We propose to deploy the system overlooking the main pedestrian entrance of City University of Hong Kong, which connects the university to the Festival Walk shopping mall. The performance of the surveillance system will be evaluated over a 2-year timeframe. This long-term experiment will provide key insights on the reliability of the system in various environmental conditions, and seed future research on modeling the dynamics of crowd flow over long periods of time. The development and deployment of a reliable crowd counting system will improve the understanding of crowd flow within large urban environments, which will have a broad long-term impact on the management and planning of the city’s resources.


Project number9041552
Grant typeGRF
Effective start/end date1/01/1129/05/15