Research On Key Technologies Of UAV Multidimensional Intelligent Recognition

With the rapid growth of Unmanned Aerial Vehicle(UAV)market and the continuous increase of purchase volume,the "black flight" incidents of UAV flying without license and flying without permission happen repeatedly,which bring serious hidden dangers to aviation safety,national safety,personal safety and public safety.Especially in the square,park and other areas of civil public life,how to counter the threat of "black flight" UAV has become an urgent problem.Identifying UAV is a key and difficult problem in UAV threat control technology.Aiming at the problem of how to effectively and efficiently identify UAV in square,park and other civil public living areas,this thesis studies the multi-dimensional intelligent identification method.Firstly,the voice method is used to recognize the UAV,and a method of UAV voice recognition based on convolution neural network is proposed,and experiments are carried out.The experimental results prove the feasibility of UAV voice recognition.Secondly,the image method is used to recognize the UAV,and a method of image recognition based on convolution neural network is proposed.The image of UAV is processed and the image is recognized by neural network.The experiment proves that the image recognition of UAV is effective and feasible.Then,the UAV is identified by radio,and a method of UAV identification based on convolution neural network is proposed.The radio signal is transformed into a spectrum map through discrete Fourier transform,and then into a spectrum waterfall map.The spectrum waterfall map contains the characteristics of the UAV signal,and then enters the convolution neural network for identification.The experiment proves the feasibility of radio identification of UAV.Finally,three kinds of UAV recognition methods are fused,and the multi-dimensional intelligent recognition method of UAV is studied,and a multi-dimensional intelligent recognition system of UAV is designed.Through experiments,the feasibility and effectiveness of this method are verified.