The Road Tunnel Flame Detection Based On The Video

In recent years, the number of vehicles increased dramatically, as well as the roads and urban traffic tunnel mileage extend fast in China. The challenges of the tunnel fire safety are enormous. Traditional smoke sense or temperature sense flame detectors are more applicable to the air flow is small and indoor space. Alarm will be issued from the smoke sensor or temperature sensor, when smoke gathered in a certain concentration or temperature reaches a certain degree. It is not a timely warning. Early detection of road tunnel fire is the current important research directions. Video-based flame detection can have visibility and accurately determine the location of the ignition point; Video-based flame detection can discover the tunnel fire more early than the traditional detection. The work completed in this article is:in the camera monitoring area of the road tunnel, if a fire occurs, flame detection procedures use image processing algorithms to detect out the flame, and then give the alarm information.In this thesis, capture video with analog cameras in road tunnel, validate a variety of flame detection algorithm on a computer, select the flame detection algorithm for road tunnel which is able to distinguish between the fire and tunnel lighting, vehicle lighting. Flame detection procedures use the motion feature, color feature, rectangle fit factor, wavelet energy and movement cumulative nature for effective detection of fire in images.In the video image segmentation:This thesis studies the motion detection method used in flame detection, and the thesis studies the color detection method used in flame detection. Analyzes and compares the result of the parallel execution and sequential execution of movement detection and color judgment. According the tunnel environment, the thesis selects out the real-time and accurate algorithm from a variety of algorithms.In the feature extraction of the flame:To find the difference of the movement cumulative between the flame and cars light inside the tunnel, the different of the rectangle fit factor between the flame and cars light, the difference of the wavelet energy between the ordinary objects and the flame by doing experiments. The movement detection method can exclude the tunnel lights. When movement detection and color detection run at the same time, the procedures can exclude part of the impact of cars light, according the difference of the rectangle fit factor, the wavelet energy and the movement cumulative between the car light and the fires, the procedures can exclude the impact of cars light and other interfering substances. In the camera monitoring area of the road tunnel, if a fire occurs, flame detection procedures can detect out the flame.