Smoke Detection in Video using a Raspberry Pi

Many different algorithms have been developed for smoke detection of in video. However, most of these algorithms have been implemented in simulations with specialized software or computers with high computational capacity. This article presents results of implementation of a method to smoke detection in video, which was implemented in a Raspberry Pi embedded computer using OpenCV library. This method proposes to use different smoke characteristics such as: movement, color, direction, growth, accumulation and a background subtraction method, so that movement detection is as optimal as possible to be able to detect early smoke presence and prevent a possible mishap. This algorithm was evaluated using 22 different videos obtaining up to 90.9% accuracy in smoke detection and up to 9.1% false positives, showing that proposed method obtains favorable results.