AUTOMATIC FIRE DETECTION SYSTEM

Abstract: The threat because of fire has become increasingly serious to people’s lives and property. To overcome the problem of traditional fire detection i.e., false alarm rate, we proposed an innovative detection method based on multi feature fusion of flame. First, it is considered as fire pre-processing stage where we combined the motion detection and colour detection of the flame. Second, flame has a certain similarity in the sequence of the image, even though it is irregular. Then, we included some features extraction to improve the accuracy of recognition. The new features are of spatial variability, shape variability, and area variability of the flame. At the end, we used a support vector machine for training and completed the analysis of candidate fire image. Thus, achieved automatic fire monitoring. The method can also be applied to real-time camera monitoring systems, like home security, forest fire alarms, and commercial monitoring.

Keywords: flame, camera, fire detection, alarm, monitor, SVM, fusion.