Journal of Multidisciplinary Studies Vol. 9, Issue No. 2, pp. 18-36, December 2020 ISSN 2350-7020 (Print) ISSN 2362-9436 (Online) doi: http://dx.doi.org/10.7828/jmds.v9i2.1484 18 Indoor Video-Based Smoke Detection using Gaussian Mixture Model and Motion-based Tracking John Rommel Y. Pedros, Roseclaremath A. Caroro, Manolo R. Jumalon Bren S. Cajeta, Daniel V. Renacia College of Computer Studies, Misamis University, Ozamiz City, Philippines Corresponding author: Roseclaremath A. Caroro, email: claire130705@gmail.com Abstract Smoke is the leading cause of death due to suffocation as fire emits smoke earlier than other signatures throughout fire growth and development stages. Thus, its rapid detection can maximize the probability of successful fire suppression and survivability. Traditional methods detect smoke but are inefficient when under certain circumstances. However, video-based smoke detection is increasingly popular, although most did not study its dynamic characteristics such as its motion, speed, and environmental factors. This study presented a method for indoor video-based smoke detection composed of a static detection of foreground or moving pixels using GMM and the dynamic detection through motion object tracking using Kalman Filter to verify and analyze the smoke behavior. The results showed that the algorithm detected smoke effectively given varied test circumstances. Although, it also detects non-smoke objects since the algorithm focuses on detecting moving objects. This study contributes an algorithm for developers working on alarm systems and similar works. Keywords: smoke detection algorithm, motion-based tracking, GMM, Kalman filter, foreground detection