Real-time wildfire smoke detection on moving camera

Early detection of forest fires has a huge importance in terms of the forest protection. In previous studies, wildfires detected with the help of fixed camera and computer. The camera used only monitors a specific region, for the rest of this region either an another camera is used or it just cannot be monitored. In this paper, we propose an approach that can monitor the whole 360^o field using only a single camera for the early detection of a potential wildfire by detecting presence of smoke. In this work, motion information of the camera is calculated from the frames that are recevied from the camera. By using this movement information, a fixed background image is obtained and updated at certain time periods. Using this background image, the current images are examined and possible occurences of smoke in the video is detected and presented in this paper.