Design And Construction Of Fire Detection Platform For Uav

Fire is a hazard to human life and property.It has been committed to research and development of such a major disaster forecasting system, especially in the complex environment.Fire detection system with low false alarm caused widespread concern everyone.Many experts and scholars were attracted to Vision-based fire detection methods because of its wide range of applications. UAV has wide applications, fire detection using aerial robot as the platform is a challenging cutting-edge multidisciplinary research project in the world which has attracted many research groups.This paper present a design of UAV fire detection system, including UAV hardware platforms, flame detection algorithm, the three-dimensional reconstruction and tracking of the fire point, data fusion on BP neural network.In the first chapter,the paper intruduced the hardware structure of the platform, the form of image collection,images-capture using the Video4Linux2 API the application programming;Images compression using open source libraries FFMPEG; Wireless transmission of images using streaming media library Live555 RTSP protocol.Then, after introduces fire detection and UAV technology at home and abroad.We compared the advantages and disadvantages of various technologies, color camera-based fire detection algorithm and the infrared-camera-based flame detection algorithm.,We combined the RGB fire criterion and HSI fire criterion in the color camera flame detection algorithm; infrared camera detection algorithm is proposed to use slope threshold of the flame temperature to detect high-risk point of fire and using flame dynamics method to detect fire flame.Using three-dimensional computer vision-based method, we reconstruct high-risk point of fire or fire point.The images from two cameras can be transmitted back to the screen on the ground and delineate the ignition point location and guide the MCU turntable automatic tracking of the point.Finally, using three color components R, G, B and infrared images of the gray value as intput of BP neural network,training the neural network to judge the authenticity of fire information,in order to improve the reliability of the system and reduce false alarm.