This paper presents a real-time imaging acquisition and processing system for early smoke detection in indoor scenarios such as home/office, warehouse, industrial building, or on-board vehicles (train, bus or ship.). The main idea is integrating standard gas/temperature sensors with the results of a video-based smoke detection processing exploiting its temporal, spatial and chromatic characteristics.
The proposed technique uses common video surveillance cameras, operating in visible spectrum, that are available at low cost or are already installed in the indoor scenario. The proposed algorithm has been successfully verified with a large set of test videos in indoor scenario and implemented in real-time in a low-cost DSP-based platform with power consumption of about 100 mW and a board size of 10 cm. The proposed system features a detection latency of few seconds while traditional systems require some minutes.