A Proper Orthogonal Decomposition based algorithm for smoke filtering in videos

Wildfires exhibit threats of all magnitudes and types to life and property. Past records suggest inevitable need of complete situational awareness and importance of the use of Unmanned Aerial Systems (UAS) to improve the wildland fire management by using onboard digital cameras. A major issue is the presence of smoke that occludes the hot spots in videos taken from such cameras. This research work focuses on reconstructing images from video of scenes occluded by thick smoke and a method for filtering smoke occlusions in fire image streams using Proper Orthogonal Decomposition (POD). Assuming that the image of the wildfire is taken from a static camera, the smoke will be moving over a stream of images or a video but the background will be static. Using POD, the smoke is filtered out of the video and clear background with fire can be seen in the output images. It provides an efficient way of capturing the dominant components of an infinite-dimensional process with only a finite number of “modes”. The technique is applied to a number of sample videos and it is demonstrated at the smoke is sufficiently removed from the video with the background information intact.