Abstract: Video fire detection has important practical significance for fire safety. According to low
detection rate and high false-alarm rate in the current flame detection algorithm, a novel method to
detect flame is proposed, which combined flame motion detection, color clues and realized the
multi-feature fusion based on the analytic hierarchy process (AHP). Firstly, the background modeling
is carried out by optimizing the selective background updating to extract the fire-like moving region.
Secondly, the color analysis technique is applied to extract suspiciousflametarget. Thirdly, study the
dynamic features of flame and discuss detailed frequency property, angle, roundness characteristics,
areagrowth and overall movement of the flame. Finally, a method of realizing the multi-feature fusion
based on AHP for calculating the weight of the dynamic features of flame is proposed. The
experimental results on the flame video library show that the proposed method is accurate and robust,
which has high detection rate and low false-alarm rate.

Key words: detection, background updating, color detection, dynamic features, analytic hierarchy
process