An Early Fire Image Detection and Identification Algorithm Based on DFBIR Model

An algorithm of early fire image detection and identification based on discrete fractal Brownian incremental random field model (DFBIR) is proposed in this paper. The flame color brightness and motion clues are chosen as the imprint of the fire image. At first, the suspicious fire region is detected by the flame color model. And the differential model is used to analyze if the suspicious fire region extends. If the suspicious region extends, the early fire image is further detected and identified by the algorithm of DFBIR model. The algorithm can exclude the mendacious fire and detect the real fire in three steps as follows. First, the size of the rectangular window is confirmed. Second, the fractal dimension and the fractal deviation are computed. Third, the migratory window calculation is done to make a final decision. The experimental results show that the algorithm can successfully detect and identify fire.