[M-GI35] Earth and planetary informatics with huge data management

[MGI35-03] Smoke detection based on optical flow using the variance of multiple frames of video

*Kazutaka Kikuta¹, Yuki MURAKAMI¹, Ken T. Murata¹ (1.National Institute of Information and Communications Technology)

Keywords:Visual IoT, Optical flow, Automatic detection

We propose a system that detects smoke using video IoT technology for disaster prevention. This system detects smoke in the PTZ (Pan-tilt-zoom) camera videos installed outdoors from its movement, then moves and zooms the camera's viewpoint to the position of the smoke, and acquires the enlarged smoke image. Smoke detection leads to the early detection of fire, which is important for disaster prevention. This method uses optical flow to detect the movement of smoke. Video High-resolution and high-time-resolution video of IoT is suitable for applying optical flow. Optical flow is a method of detecting the movements of objects from the luminance information between image frames, but when applied to outdoor monitoring images, it detects moving objects such as cars other than smoke. In this research, optical flow is used to evaluate the variance of the motion vector of multiple video frames to reduce noise other than smoke and perform accurate smoke detection. This utilizes the characteristic that the area and the motion direction of smoke are almost constant in the videos for several seconds. By operating the PTZ camera based on the position of the acquired smoke, it is possible to acquire higher-definition smoke images in real-time. In this study, we introduce a proposed method for a real-time monitoring system with an outdoor camera and demonstrate the smoke detection.