Software Design Of Smoke Recognition For Forest Fire Based On Multi-channel Video Images

Forest fire is the most serious threat facing forest resources,causing great damage to human production and life and the ecological environment of nature.Therefore,early warning of forest fires for humans and nature is of great significance.With the development of video surveillance technology,video-based forest fire detection technology has received widespread attention.Compared with the traditional forest fire prevention technology,this technology has the advantages of fast response,high accuracy and intelligence,which is very suitable for fire warning and monitoring in a large area such as forests.Because it’s difficult to directly observe an open fire in early forest fires,it is often accompanied by smoldering objects such as dead branches and leaves in the forest that produce a lot of smoke.Therefore,forest fire video image smoke recognition is the key to detect early forest fires.Based on this feature,this paper designs forest fire smoke recognition software based on multi-channel video images,which aims to encourage the development of forest fire prevention technology towards high efficiency and intelligence.The main research contents are as follows:(1)A forest fire video images smoke recognition algorithm based on PoolNet saliency detection and SURF-ViBe motion detection model is proposed.Firstly,preprocessing the video image with such as median filtering,color processing and denoising.Then using the deep learning method based on PoolNet saliency detection and SURFViBe motion model to extract the video image suspected smoke area.Then extracting the dynamic and the static characteristics of smoke,such as the absolute smoke area growth and diffusion of the suspected smoke area,and the static characteristics of smoke such as background blur and texture.Finally,SVM is used to train the extracted smoke features to achieve the classification and recognition effect of smoke.The experiment and test results show that the accuracy rate of the smoke recognition algorithm in this paper is more than 90 %,compared with the traditional smoke recognition algorithm,the recognition rate and real-time performance are improved.(2)A forest fire smoke recognition software based on multi-channel video images is designed and implemented.Firstly,the overall software framework is designed,and the software functions,algorithms and performance requirements are analyzed.Then,based on the research of smoke recognition algorithm,a modular programming idea is used to design the forest fire video images smoke recognition software in detail,module functions such as device login and management,multi-channel video data collection,multi-channel video images smoke recognition,multi-channel device linkage,monitoring and automatic cruise,manual device control and fire alarm information database storage are realized.The use of CUDA-based GPU parallel acceleration technology can effectively solve the problems of large data volume and low real-time performance caused by simultaneous processing of multi-channel video data.The multi-channel equipment linkage function can effectively expand the monitoring range,improve the accuracy rate of smoke recognition,and at the same time it is convenient for the commander to remotely understand the fire situation,formulate a fire fighting plan and direct fire fighting.Finally,the actual functions of the software are verified by actual tests and the established requirements are met.