A New Hybrid Feature Extraction Method Based on Accurate Motion Area

Abstract

As the effect of fire detection can be severely affected by the following two problems: the extraction of moving target area with low accuracy and fire detection with a single feature, this paper puts forward a new method of moving target detection, then analyzes and extracts the features of flame, such as color, energy, flashing, and contour features. So whether there is a fire occur in the motion area can be detected exactly. Experimental results show that the proposed fire detection method can reduce the error rate, improve the real-time and anti-interference ability.