Design Of Real-Time Detection And Tracking Of Anti-Drone System Based On Sky Background

With the increase of illegal flight,which seriously threaten the safety of civil aviation at airports,the development of anti-drone systems has become increasingly urgent.In particular,the development of reconnaissance and detection subsystems in such systems is more critical and complex.At present,such subsystems generally have disadvantages such as low tracking accuracy,poor real-time performance,poor adaptability,and high hardware cost.Aiming at the civil airport anti-drone project of the research group,this thesis designs a subsystem that uses "FPGA+GPU" as the main processor,which can detect and track the target drone in the visible light image in real time.This subsystem is combined with civil radar to realize compound detection,realizing real-time detection of illegal flying of drones under a simple sky background,creating conditions for drone countermeasures.The main contents of this thesis are as follows:(1)By analyzing the development status of anti-drone systems at home and abroad,and comprehensively comparing the architecture characteristics of existing detection and tracking systems,the overall design scheme of this thesis is proposed.The main content includes determining the use of the "FPGA+GPU" system architecture,using FPGA as the core processor to complete image acquisition and preprocessing,and using GPU as the core processor to complete the detection and tracking algorithm implementation and logic control.(2)Combining the technical indicators and functional requirements,this thesis designs the overall implementation architecture of the system.The architecture includes determining the system hardware architecture design,determining the selection of the core processor,determining the communication interface,designing the preprocessing algorithm,determining the selection and feasibility analysis of the detection and tracking algorithm,and determining the system software framework,and so on.(3)This thesis completes the circuit planning and design of each functional module.The main content includes completing the circuit schematic design and PCB design of the corresponding modules,and verifying the functions and interfaces of each circuit board to ensure that the built system hardware platform can operate stably and reliably for a long time in actual work.(4)This thesis completes the logic and code design of each functional module.The main content includes the logic and code design of image acquisition,preprocessing,reading and writing of DDR3,and the use of optical fiber for data transmission and reception on FPGA,the logic and code design of image data interaction and instruction transmission between FPGA and GPU,the logic and code design of the software control and the implementation of the selected algorithm on GPU,the logic and code design of the simulation training by inputting the images of Camera Link simulation training image,and so on.(5)This thesis completes the system function test and verification.The main content includes the independent function test of each module,the test of the image data transmission and command interaction function between the modules,and the final system test through the flying target aircraft,and so on.This thesis repeatedly debugs the code logic to ensure that the system meets the technical indicators and functional requirements.Based on the novel design concept and system architecture,the system developed in this thesis is effectively applied to the civil airport anti-drone scene.The measured1920×1200@50Hz video image of the drone has the longest effective tracking distance of2.5km in the small field of view,and 1.5km in the large field of view.The overall delay of the system from imaging to GPU network display is controlled within 50 ms,which realizes the high-precision real-time detection and tracking function of the drone,especially the realtime performance of the system basically achieves the domestic leading level.From the beginning of image collection to the completion of GPU network display,the overall delay of this system is controlled within 50 ms,which realizes the real-time detection and tracking function of drone,especially the real-time performance of the system basically achieves the domestic leading level.Therefore,the system developed in this thesis has high engineering application value and a high reference value for the research and development of similar equipment.