Research on Flame Image Detection

Abstract

Fire flame detection can reduce the damage of people’s life and property in maximum. We analysis the significant characteristics such as color feature, sharp corners, flicker frequency, change of area, flame core features such as the relative area in the suspected areas as fire classification. Through the characteristics of the fire video image analysis, first, select color features to pre-judge, then couple with area change rate and sharp corners to determine the false fire target, last, adapt the algorithm to tall space, outdoor environment and some fire detection under complicated environment. Comparing with traditional detection, the algorithm can find the fire accurately and timely, reduce the amount of calculation, improves fire recognition accuracy, and reduce false positives omission phenomenon.