Smoke Detection from Video Images


M. Palaniswami, Slaven Marusic;
Post Doctoral Research Fellow: Jayavardhana Gubbi.

Student: Sutharshan Rajasegarar.
Introduction: Visual sensor network is becoming a reality with increased computational capabilities. This includes an array of cameras positioned strategically to monitor sensitive areas.
Significance: Of late, there has been an increased interest in detection of fire in critical infrastructures including tunnels, hangers and also in detection of bush fires. Due to the deployment of visual sensor networks for surveillance purposes, there is a possibility of using this existing infrastructure in designing early warning system. This can be accomplished by the detection of smoke. When continuous video is being analysed, if there is a fire, smoke will cause the scene to appear semi transparent. This will reduce clarity and hence there is a reduction in high frequency component.
Applications: This project proposes to make use of the above feature in designing an early warning system. Multi camera Intelligent Surveillance systems are one of the key offerings of iOmniscient. iOmniscient seek low-cost and highspeed algorithms that can run in the distributed video surveillance network to offer smarter video surveillance solutions. Their Non-Moving Object Detector won us accolades all over the world and we are keen to extend it to other smart features like detection of slip and fall, tracking objects across multiple camera systems and also explore the application of multiple, heterogeneous sensors to aid the video surveillance solutions.
Challenges: This collaborative project under the purview of DEST-International Science Linkage grant on Distributed Sensor Networks intends to develop new algorithms for smoke detection. This falls under the category of surveillance in visual sensor networks of DESTISL project.
top of page