Research Of Fire Smoke Detection Based On Weighted Feature Combination Method

Fire may bring disaster to human beings and cause a loss to a country and its people.Therefore, fire detection plays an important role in protecting human lives and ensuringproperty security. The traditional fire detection technology needs to lay special sensors, whichis difficult to implement in modern high buildings and is not cost-effective. It is urgent todevelop new fire detection technologies. With the wide use of intelligent monitoringequipment, the technology of fire detection based on image processing emerges to satisfy therequirement for new fire detection technologies. The smoke in fire is an important and usefulsignal. We use the existing monitoring equipment to receive video sequences, then adopt themethod of image processing to detect the smoke in video images, and fially fulfill the goal ofthe fire alarm.This paper mainly researches the method of early fire smoke identification based onimage processing. First, we detect the moving object by gaussian background modeling andbackground subtraction and extract the suspicious smoke area with the connected componentlabeling method to narrow the testing scope and improve the detection efficiency. Then weextract the four characteristics of smoke as follows: the color feature, the irregular contour, thediffusion area, the main upward movement direction. Next, using the voting combinationclassifier to the four smoke features, we obtain the final recognition result. The experimentshows that the method used in this paper can eliminate noise effectively and identify thesmoke accurately. Compared with other methods, it has higher detection accuracy, lower errorrate, and better application. The article also summarizes the work done for this paper, and putsforward the future research direction for the inadequacies.