Research On UAV Detection And Recognition Technology Based On Intelligent Optimization

The rapid development of UAV technology makes UAV more intelligent and popular.However,the security problems brought by UAVs have also been concerned by the society.Controlling UAVs is an important means to solve the security problems,and the premise of counteracting UAVs is the identification of UAVs.UAV recognition faces many problems,such as multiple UAV models,variable signal types and low recognition accuracy.In view of this problem,based on the intelligent optimization idea,this thesis uses time-frequency image recognition and RF-DNA method to study the detection and recognition of UAV.The specific work contents are as follows:Firstly,the structure characteristics and performance parameters of UAV Communication System and frequency hopping communication system are studied.Based on the current time-frequency analysis technology,the frequency hopping signal is simulated and compared,and the time-frequency analysis method suitable for this thesis is selected.The selected time-frequency analysis method is used to obtain the UAV time-frequency image,and then the time-frequency image is enhanced to improve the generalization ability of the model.Two feature extraction methods,hog and LBP,are used to extract time-frequency image features.To solve the problem of low recognition rate of single feature extraction method,this thesis proposes to combine hog and LBP features to form a feature database.The SVM multi classifier is designed to classify and recognize UAV.The experiment proves the effectiveness of this method.Secondly,in view of the low efficiency of manual feature extraction by hog and LBP,a convolutional neural network is proposed to automatically extract and recognize image features.First,the network model inception-resnetv2 is adjusted for image feature extraction,then the UAV signal time-domain sequence is converted into one-dimensional frequency-domain sequence,and the network model is designed for feature extraction of the one-dimensional frequency-domain sequence.Finally,the two features are combined and the method is applied to classify and recognize the UAV.This method is efficient and automatic.Finally,in order to further optimize the precision of UAV type identification,a method of UAV radio frequency fingerprint identification based on transient characteristics is proposed.Three kinds of feature data are used for feature extraction of UAV transient part:instantaneous frequency,instantaneous phase and instantaneous amplitude.Dimension reduction of feature data based on PCA to filter the interference of external information.Four groups of comparative experiments were set up by using feature dimension and standard deviation.The experimental results show that UAVs of the same type and different individuals can be effectively identified.