Autonomous Aerial Fire Extinguisher is one of the categories in Indonesian Aerial Robot Contest (Kontes Robot Terbang Indonesia-KRTI). This is a national contest for Unmanned Aerial Vehicle (UAV). The background of this contest is to give solution for forest fire in Indonesia. In this contest, the UAV needed to detect and determine a fire position placed in an indoor area which then will be extinguished. The most accurate UAV managed to extinguish the fire will be crowned as the champion. To support the detection process, we need to find the best fire detection algorithm to be implemented in the autonomous quadcopter. By analyzing the color space from RGB, we got the average accuracy on real time flight testing is of 72,94%. This value is fit for the contest condition which required minimum accuracy is 70%. Thus, this paper propose a new method that can be implemented in real forest fire detection.