Research And Realization Of Forest Fire Detection Technology Based On Video Sequence

Intelligent video processing technology for forest fires is a new field,which is different from indoor fire detection.Because of the wide range of forest environment,the distance of the fires is often far away;because of the uncertainty of the natural environment,there are many disturbing substances in the forest;because of the changes of seasonal and weather,the form of the forest itself is varied.This makes the forest fire detection a challenging task.Therefore,the study of forest fire detection technology based on video sequence has great theoretical significance and practical value.This paper launches the study around the detection of fire and smoke in forest.Mainly including the study of forest fire detection algorithm,the relevant program simulation,software development and embedded implementation.Firstly,the basic knowledge of video image processing is introduced,including the basic principles of color model,video representation,moving object detection,region connectivity,wavelet analysis and super pixel segmentation.Then,in order to extract the fire and smoke target in the forest fire scene,the static and dynamic characteristics of the fire are analyzed.The suitable feature is extracted and the algorithm flow is formed.Because the robustness of the pixel-level fire detection algorithm is not strong,this paper has done some research on the forest smoke detection by super-pixel segmentation.The advantage of the super-pixel segmentation is that the detection result is a continuous region,which avoids the debris region.And can eliminate the influence of the noise point on the robustness of the detection system.SVM is used to classify and identify the super-pixel blocks,which avoids the cumbersome step of adjusting the threshold value.Finally,the forest fire detection platform based on MFC is developed to meet the needs of practical engineering.What's more,the forest fire detection algorithm is developed on DM8168 development board.The experimental results show that the forest fire algorithm based on video sequence has high recognition rate,and the fire detection based on image segmentation has the significance of further research.The fire detection platform based on DM8168 has certain practical value.