Design And Development Of Smoke Detection System For Forest Fire

Forest fire is one of the natural disasters that seriously threaten the lives and properties of human beings.Early forest fire detection technology is basing on sensors,which has the disadvantages of slow response time and large-scale deployment costs.While real-time video surveillance for fire detection has wide application scope and fast response speed.In the early stages of fire,it is often accompanied by a large amount of smoke.Therefore,forest fire detection for real-time video is essentially detecting smoke images in video.However,in the process of smoke detection,there are many factors such as interferences and illumination changes reduce the recognition rate of smoke.Therefore,the smoke detection system of forest fires is designed and developed in this paper.The main contents of this thesis are as follows:(1)The integrity of the foreground extraction directly affects the accuracy of subsequent smoke detection.In this paper,common foreground extraction algorithms are discussed and an adaptive background update algorithm is selected as a method for extracting the foreground of the system.On this basis,the background update parameters are improved to obtain a novel adaptive background update algorithm,which will be used for foreground extraction.However,due to the influence of light changes on the smoke detection,the lighting effects removed with normalization from the background frame and the current frame obtained from the background update model.Then the foreground of the image is extracted using the background difference method.For non-smoke objects in the foreground image,the RGB color model is used to remove some non-smoke objects in the foreground image.Finally the candidate smoke is obtained.(2)Interferences are difficulties in smoke recognition.There are various types of interferences in the smoke image,including moving vehicles,shaking branches,and the moving objects which have the similar color with smoke.For the removal of interfering objects,the method of combining the static and dynamic features of smoke is generally used.For the object which has similar color with smoke,according to the dynamic characteristics of smoke and LBP feature to extract texture features from the image.Training positive and negative samples and LBP texture images using SVM,and the judgment is made based on the image information contained in the image feature vector.(3)Through the research of the smoke detection algorithm,the forest fire detection system is designed and implemented in this paper,which is divided into three modules:image preprocessing module,foreground extraction module,feature detection and identification module.Then,using MFC to implement the system interface and test the function and real-time performance of the entire system.