Research On Forest Fire Monitoring System Based On Unmanned Aerial Vehicle

Forest fire has the characteristics of large disaster area and high temperature in the fire area.Once the forest fire occurs,it is difficult to control,causing serious damage to the natural environment and the lives and properties of related personnel.Therefore,effective forest monitoring is very important.Traditional monitoring methods such as ground patrol,aviation patrol and satellite remote sensing,not only have high cost of personnel and equipment,but also are greatly affected by the terrain,and there are blind spots in the monitoring process,which has great limitations.With the development of electronic technology,UAV,as a new type of aviation platform,has the characteristics of high mobility and convenient operation.Its application in the field of forest fire monitoring has also attracted more and more attention.In this paper,forest fire monitoring is taken as the goal and four rotor UAV as the basic platform,the system structure,cruise control algorithm,and image detection and recognition algorithm of forest fire monitoring UAV are studied.A forest fire monitoring system based on UAV is developed,which can achieve the functions of forest fire image acquisition,analysis,processing and fire early warning.The main research contents include:(1)The flight principle and dynamic model of the forest fire monitoring drone are studied.In order to reduce the blind spots and complete the high-precision cruise mission,a drone integral sliding mode controller is designed.The weight of the equipment and the impact caused by the weather such as high winds,use a non-linear observer to observe the state and interference of the forest fire monitoring drone during cruise,and enhance the anti-jamming resistance of the forest detection drone,and then designed and developed Software and hardware systems for forest fire monitoring drones and image transmission systems.(2)The target region extraction and feature extraction methods of forest fire and smoke are studied.By analyzing the color distribution of forest flame in RGB color space and smoke in YUV color space,a segmentation method of forest flame and smoke based on color is proposed.The image features of forest flame and smoke are studied from the shape,texture and motion change of forest flame and smoke,and the feature vectors of forest flame and smoke are extracted for the following image classification processing.(3)The classification method of forest fire image based on particle swarm optimization and kernel limit learning machine is studied.In order to improve the classification effect of the kernel limit learning machine,particle swarm optimization algorithm is used to optimize the parameters of the kernel limit learning machine.A forest fire image classifier based on improved PSO-KELM is proposed to classify forest fire image and non forest fire image,and the accuracy and efficiency of the proposed classification algorithm are verified by using the video of forest fire.QT creator and Open CV are used to develop ground monitoring software for real-time forest monitoring.