The Research Of Automatic Image Recognition Technology Based On Forest Fire Video Monitoring

Forest fire is of devastating natural disaster.Forest fires break out hundreds of thousands of times every year throughout the world,which seriously endangers the earth's ecological resources and the survival and development of mankind.Therefor,prevention and surveillance of forest fires have been focusing by researchers around the world.With the rapid development of science and technology,a great variety of emerging technical means are gradually applied to the surveillance of forest fire,for example,satellite remote sensing,intelligent sensor,unmanned aerial vehicle,infrared detection,etc.Traditional manual patrol and the means of observation have also changed dramatically.Because of the wide range of application and economical efficiency,one of the new fire monitoring and recognition technologies based on video image processing technology is being watched by more and more researchers.Research topic of the thesis is forest fire video image monitoring and identification system based on image and pattern recognition technology.The purpose of the system is to achieve a combination of video surveillance hardware system installed in the forest area and fire identification software system,and can carry out all-weather forest fire monitoring and identification.The research topic is of innovative and practical significance.The thesis mainly elaborates research content from three aspects.First,forest fire monitoring and prevention technology development status are analyzed and compared.The basic theory of image processing and pattern recognition is summarized;second,functional requirements and system structure of forest fire video image monitoring and identification system are analyzed and designed.The system is reached mainly through technical processes including the image preprocessing,moving target detection,target image feature extraction and the technology of classifier identification.In addition,by changing the identification judgment section into the smoke and flame dual module classifier,all-weather monitoring requirements of the system are provided.In the system implementation method,the median filter and gray scale conversion techniques are used to carry out pre-processing of image denoising and simplifing.Next the hybrid Gaussian background difference technique is used to detect the moving target in the video background.And then,color,shape,texture and dynamic characteristics from the detected target image are extracted and a feature vector is fused.At last,the technical thinking is identified and judged by support vector machine(SVM)classifier.Third,according to the design and principle of the system structure,the study and improvement of the algorithm are realized and the results of the experiment are analyzed.Finally,through the experiment,the systems will achieve the moving target detection,feature extraction and recognition function of the sample video.Through the optimization of the algorithm,the complexity of the operation will reduce to a certain extent and the detection efficiency and accuracy will improve.The result will further verify the feasibility and practicability of the system design.