Study On Forest Fire Detection And Information Acquisition Based On Video Image

Forest is the most precious wealth given to human by the earth,which is closely related to the survival of human beings and the development of social economy.However,in recent years,the forest is vulnerable to the invasion and destruction of fire.Once the forest is attacked by fire,it will cause huge losses,such as causing casualties,reducing species diversity,breaking the balance of the ecosystem and so on.It can be seen that the prevention and control of forest fire is imminent and has a very far-reaching practical significance.However,the traditional building fire detector,manual inspection and observation tower detection methods have corresponding shortcomings,which are not suitable for open forest areas.Therefore,in order to detect and control the fire in time,it is very important to study the reliable forest fire recognition method and obtain the fire information in real time by image processing.Based on the above situation,this paper mainly adopts the method of combining model experimental and simulation,and analyzes the forest fire detection method and real-time acquisition of fire information based on image processing technology:(1)This paper summarizes and analyzes the shortcomings of some existing fire suspected area extraction algorithms,and builds a fire suspected area extraction algorithm based on super-pixel segmentation and color model.The texture and color features of forest fire image are studied by image processing.(2)The small-scale experimental platform of forest fire model is designed and the forest fire model experiments are carried out,which provides materials for the subsequent forest fire detection data set and forest fire information acquisition.The special fire data set of forest fire with rich scenes and small repeatability is established by collecting and sorting out the experimental materials and other database resources.(3)Based on Python programming language,two models of SVM(Support Vector Machine)and Adaboost(Adaptive Boosting)are established for forest fire detection.On this basis,the Adaboost-SVM strong classifier is built for forest fire detection and the small program of forest fire detection is designed and developed.Finally,the accuracy of forest fire identification of the three models is compared and analyzed,the results of Python simulation show that the accuracy of Ada Boost-SVM is the best,followed by Ada Boost,and SVM is the worst.(4)Based on video images,the algorithm of obtaining flame height information is improved by python programming language,and the algorithms of obtaining information such as flame area,shape,fire line intensity and spreading speed are designed,which provides a solid foundation for obtaining fire information.Taking the forest fire model experiments as the material,the above key fire information(flame height,area,shape,fire line intensity and spreading speed)is obtained in real time through image processing and analyzed,which provides reference and technical support for the prediction of fire development trend,so that firefighters can timely grasp the fire situation,control the fire situation and improve the fire fighting and rescue ability.