Research On Fire Image Detection Algorithms Based On ViBe

As one of the common disasters in electric power production,fire threaten the equipment and people,even stops the power system to work stably.So,how to recognize fire and take measures to it becomes the research purpose of fire prevention and control.Traditional induction fire detectors perceive flame and smoke through temperature,air particle content,sound and light.Such detection method is easily disturbed by other environmental factors,and causes false alarm due to smoking and other problems.Faced with such problems,this paper designs a fast fire detection method based on image processing and optimization algorithm.This paper focus on the research of a host-based fire image detection system.Under the untouched condition,the scene of substation is collected to the monitoring system by the camera,then the flame is recognized and located through the flame detection program algorithm based on open-CV.Dynamic detection provides a lot for image processing,and flame recognition is a strangely part of dynamic detection.In this paper,based on the design of monitoring application system,the algorithm of flame recognition and measurement is introduced in detail.Firstly,the color is between white and black.Secondly,the flame edge appears through Canny edge detector.With it,firstly,image noise is smoothed by Gaussian filter;secondly,why the edge point is attractive is due to Canny;thirdly,non-maximum suppression is the same as former;fourthly,flame edge is detected by using double threshold method to connect the edge points,and then features are extracted by using cluster labeling.Recognize and detect the flame,locate the suspected flame position: binarize the gray color image by using the optimal threshold segmentation.Fifthly,the inverted binary image is labeled by blocks,and the required area is selected.Sixthly,using the cluster marking method,the contour of the cluster is tracked,and the results are displayed on the screen from the aspects of the color,brightness,flicker,area change rate and shape of the flame,and the trend map of the flame change is drawn.The results of experiments show that through image preprocessing,edge detection,and the combination of dynamic and static characteristics of the cluster markers with flame,continuous monitoring of suspected fire areas in a period of time can be sensitively detected whether there is a fire in the substation.