Research And Design Of Drone-borne Target Detection System Based On Jetson Nan

Unmanned aerial vehicles(UAVs)have been widely used in many fields by virtue of their unique shooting angle and flexible characteristics.At present,the research of object detection based on drone aerial images generally saves the images acquired by the airborne camera to a storage device or transmits it to a ground receiver,and executes the object detection algorithm on a local device or a cloud server.In order to improve the real-time performance and detection efficiency of object detection,and to meet the needs of UAVs for long-distance cruises,this thesis designs an airborne object detection system for UAVs based on Jetson Nano.The specific work is as follows:(1)A design scheme of drone and airborne object detection system based on Jetson Nano is proposed.The hardware of the object detection module mainly includes Jetson Nano,camera and mobile card router.(2)The existing object detection algorithms are compared and analyzed,and YOLOv4-Tiny and YOLOv5 s are mainly researched.The drone public data set is used to train and test these two algorithms on the server,and the trained models were transplanted to Jetson Nano,and the algorithms were accelerated and optimized using Tensor RT technology.Experimental results show that the detection speeds of these two algorithms are 22 FPS and 18 FPS respectively.After comprehensively comparing the accuracy and speed of the two algorithms,YOLOv5 s is selected for the system designed in this thesis.(3)Based on HUAWEI CLOUD’s Object Storage Service(OBS)and Internet of Things technology,the program design method of image acquisition and data transmission based on Jetson Nano was studied;based on Android Studio,the display and remote control terminal were designed using Java language APP software is used to display the images after object detection in real time and control the flight of UAVs to realize human-computer interaction.(4)The experimental system was built and tested.The results show that for the purpose of detecting the presence of pedestrians and cars,the average accuracy of object detection is 98.8%;for the purpose of detecting the number of pedestrians and cars,the average accuracy is 87.5%;the transmission time of control commands is 179ms;image transmission time under Wi Fi and 4G network environment is 126 ms and 210 ms respectively.The UAV airborne object detection system based on Jetson Nano designed in this thesis integrates functions such as image acquisition,object detection,and data transmission.It realizes the real-time detection of pedestrians and cars in aerial images on the airborne end of the UAV,and can use the mobile communication network to get rid of the limitation of the use distance.It can be applied to crowd monitoring in squares or scenic spots,personnel search and rescue,highway inspections,traffic flow analysis and other scenarios,and has high application value.