Energy Saving and Intelligent Surveillance System for Escalator Based on Panorama Vision
|School||Zhejiang University of Technology|
|Course||Applied Computer Technology|
|Keywords||escalator energy-saving intelligent surveillance mixture Gaussian histogram of oriented gradient optical flow tracking Kalman predictor|
With the rapid urbanization process in past decades, escalators are widely applied in many public places, such as malls, hotels, airports and stations. However, the escalator has two existing defects about energy-saving and security. It always keeps invalid working state resulting in a huge waste of energy. Moreover, suddenness hurt accidents can easily happen on it because of its moving steps.There are two detection methods for energy-saving: infrared detection and gravity sensing detection. Their common drawback is no intelligent surveillance about the passengers’unsafe behaviors. This paper proposes a system to solve the problems about escalator-energy and security based on computer vision, attempting to save energy and detect the abnormal behaviors on escalator. So, this study is concerned with four primary issues:Human body detection and segmentation. After analyzing the drawbacks of human body detection by Gaussian mixture background model and the HOG feature matching, this paper proposes HOG (Histogram of Oriented Gradient) feature matching algorithms based on Gaussian mixture foreground extraction. It can reduce false detection rate of human objects, effectively segment the adherent human objects in real-time.Human objects tracking problem. Because some human objects fail to be segmented in several video frames, the system produces the track loss of the human objects. Here this paper discusses a new solution, the optical flow tracking algorithm combined with Kalman predictor which can greatly reduce the tracking loss rate, and make special technical preparations for the following abnormal behavior detection.Energy-saving design and realization. The system customizes several concerning areas. According to the human number and change of foreground pixels at corresponding function areas, the escalator can be opened and stopped automatically to achieve energy-savings purpose.Finally, detection of abnormal behavior of passengers. By the human object tracking results and logical definitions, some danger behaviors, such as walking in the wrong way, crossing the escalator’s border, falling down unexpectedly, can be detected and the system notify the passengers and security men about the dangerous situation by some ways.The system has the advantage of easy installation, convenient maintenance, easy implementation, low cost and convenient upgrade rebuilding .Experiment results show that the human objects detection and segmentation on the escalator can be executed in real-time and energy-saving and the detection of abnormal behavior can be implemented to ensure passenger safety.