Study of smoke detection by analysis of saturation in video

This study provides a method to detect the smoke in a field by computer vision. The software of smoke extraction is developed by using the graphical programming language. In processing the video images, it utilizes the HSI model to extract statically and dynamically the pixels of the smoke with specific saturation. The other pixels that are filtered off are regarded as the backgrounds. Elimination of noises by morphological analysis in the filtered images is applied. The results are demonstrated to successfully extract the smokes in various videos, which shows that it is promising in future for our method to be integrated in the existing video security system or further in the on-site industrial safety equipment.