Flame Object Segmentation by an Improved Frame Difference Method

This paper in view of the characteristic of traditional fire alarm system in using sensor away from fire source is hard to detect the fire early, conceives of a method of camera pattern recognition, and according to using fisheye camera which brought to top as the visual sensor and in a larger scope in the early detection of fire, proposes a frame difference method based on the cross-cutting and polymerization of flame flash point´s location, which achieved flame object segmentation. By the execution of two consecutive frames´ difference in live video streaming, this method rules out the interference of moving objects according to the crossing of the flash point´s location, and then rules out part of dissociative flash point which is interference according to the polymerization of the flash point´s location, and get suspected flame area finally. By thresholding method, after getting contour and drawing contour, we get the outline of the flame. This method can effectively segment the flame from the images, and so as to provide efficient recognition method for the early warning of the fire prevention.