ONSET FIRE DETECTION IN VIDEO SEQUENCES USING REGION BASED STRUCTURE FROM MOTION FOR NON-RIGID BODIES ALGORITHM

摘要

Onset of fire detection plays a vital role in order to prevent spreading of fire to dangerous situations. Use of image processing can achieve significantly fast results in early fire detections. However, color and edge detection image processing techniques fail to accurately detect and identify events like the onset of fire in video sequences. To detect and identify fires by using video sequences we propose a technique using region based and a new three steps search based algorithm (NTSS) and structure from motion (SFM) technique. In this paper, we are proposing an algorithm known as SFM for non-rigid bodies (SFM-NRB) algorithm. In this algorithm, we calculate a randomness figure using region wise object boundary to differentiate fire from none fire events in 9 designated locations of the field of view of the camera. The result shows that the proposed algorithm is able to detect the onset of fire within 29 frames (within 1 sec) of the onset of fire. In future it can be further improved by using a high frame rate video cameras and multi sensor detection techniques.