RF Array-based Multi-target UAV Identification System

In recent years,the UAV(Unmanned Aerial Vehicles)industry has grown rapidly.While civilian consumer drones bring convenience,“black flying” and “overfly” occur frequently,which have created an urgent need for effective regulation of drones.Laws and regulations set by government departments need to be coupled with effective low-altitude drone surveillance and identification technologies to achieve effective control,but the complex urban airspace environment poses many challenges for anti-drone systems.To address this problem,this thesis designs an RF array-based UAV identification system,develops a high-speed multi-channel data system based on FPGA,and combines deep learning methods to accomplish high-precision identification in multi-target scenarios.Specifically,the main work of this thesis is as follows.(1)This thesis designs and implements the space domain filtering acquisition system.A multi-channel data acquisition system is built with the FPGA hardware platform as the control and computational core.The MUSIC algorithm and MVDR algorithm are migrated on the FPGA to achieve filtered data acquisition in different directions in the airspace.By optimizing the data transmission paths and the design of communication and cache modules,the system synchronization and real-time performance are improved.The stability of the acquisition system operation in engineering scenarios is improved by adding functions such as automatic channel selection and antenna array calibration.(2)A deep learning based multi-target UAV recognition scheme is proposed and implemented.The scheme takes advantage of the RF array and combines the RF characteristics of the secondary target to improve the recognition accuracy of the primary target.In this thesis,we make full use of the angle information supplied by the space domain filtering acquisition system to construct a targeted model based on the angle relationship between multiple targets,and introduce the SE(Squeeze-and-Excitation)attention mechanism to achieve accurate control of the multi-temporal frequency map weights in the model input.The proposed scheme has significant performance improvement over the single antenna recognition method at all distances and achieves accurate recognition result output for each individual target.(3)This thesis designs and implements a complete system for multi-target UAV identification.After completing the algorithmic flow design,the data acquisition,signal processing,space domain filtering algorithms,and model deployment are all implemented in an FPGAbased system for real-time operation.The logical correctness and functional integrity of the modules are verified,and the logic design is verified to meet timing and resource consumption requirements.The system can operate stably for a long time and has application value.