Automatic congestion detection system for underground platforms

Abstract

An automatic monitoring system is proposed in this paper for detecting overcrowding conditions in the platforms of underground train services. Whenever overcrowding is detected, the system will notify the station operators to take appropriate actions to prevent accidents, such as people falling off or being pushed onto the tracks. The system is designed to use existing closed circuit television (CCTV) cameras for acquiring images of the platforms. In order to focus on the passengers on the platform, background subtraction and update techniques are used. In addition, due to the high variation of brightness on the platforms, a variance filter is introduced to optimize the removal of background pixels. A multi-layer feed forward neural network was developed for classifying the levels of congestion. The system was tested with recorded video from the London Bridge station, and the testing results were shown to be accurate in identifying overcrowding conditions for the unique platform environment.