Browsing by Author "Dağlı, Mehmet Ali"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    A video surveillance system for detection of potentially dangerous events in underground railway stations
    (Bahçeşehir Üniversitesi Fen Bilimleri Enstitüsü, 2013-04) Dağlı, Mehmet Ali; Eroğlu Erdem, Çiğdem
    Due to the excess number of security cameras in metro stations, security personnel cannot simultaneously monitor all of the cameras effectively. Therefore, they might miss some of the potentially dangerous events that require timely action. In this thesis, a video surveillance system, which automatically detects several potentially dangerous events, and generates warning messages for the security personnel is presented. The potentially dangerous events that are handled in this thesis consist of: safety line intrusion, proximity warning, fallen object detection, prohibited zone violation and open prohibited zone entry detection. Intrusion of the safety line along the tracks or an accidental fall on the tracks can cause lethal events if necessary precautions are not taken on time. Before the detection of the potentially dangerous events, it is important to detect the moving objects in the scene. For this purpose several state-of-the-art background estimation methods have been implemented and compared. Then moving objects are detected based on the estimated background. Another event that is informative is the state of the train, which can be either approaching, stopping, or leaving the station. For example, if the train has stopped, then detection of safety line intrusion is irrelevant, since passengers will be getting on and off the train. The layout of the cameras at the pilot station has an active role in the development of the algorithms. Video streams from four cameras in the station, which have different field of views have been used for different purposes. For example, the camera used for the detection of the prohibited zone violation is different from the camera that is used for detection of the safety line intrusion. The developed system has been tested on video data obtained from a pilot station in stanbul, where no such automatic video analysis system is currently being used. The experimental results are quite promising, and they have been verified by security personnel. The system can operate real-time and can be adjusted via a graphical user interface.