Research On Maneuvering Target Detection And Tracking Algorithm Based On Machine Learnin

Recent years,due to war and commercial needs,drone has achieved unprecedented development in the military and civilian fields.More and more military and civilian activities related to drone are emerging.Drone plays an indispensable role in the intelligence investigation and aerial photography,but it also brings about problems that endanger the lives of people such as interference with navigation,drug trafficking,and drone attacks.Because of this,the research and development of drone detection and tracking technology are very important.Drone is mobile targets and often appears in clusters,which places higher requirements on traditional target detection and tracking technology.How to improve the performance of target detection and tracking algorithms have always been a problem that researchers pay close attention to.This thesis mainly studies the target detection and tracking technology.The main research work of the thesis is briefly summarized as follows:1.Aiming at poor performance of traditional target detection algorithm under low signalto-noise ratio,this thesis proposes a target detection method based on twin support vector machines.Firstly,simulation software is used to simulate the echo data of drone detected by radar,and the echo data under different low signal-to-noise ratio is obtained.On this basis,perform data preprocessing on the echo data to obtain training data and test data,and the detection method is trained by the data.Then the detection algorithm based on constant false alarm and the detection algorithm based on support vector machine are introduced to compare the performance.Experimental results show that the proposed method can not only maintain superior detection performance at high signal-to-noise ratio.At the same time,accurate target detection can also be achieved under low signal-to-noise ratio,which solves the problem of detection accuracy decline of traditional detection methods under low signal-to-noise ratio.Finally,the high precision detection of drone is realized.2.Aiming at poor performance of traditional target tracking algorithm when the target has step maneuver,this thesis proposes an improved interactive multi-model algorithm based on back propagation neural network and introduces support vector regression network to further improve the algorithm.Firstly,use simulation software to simulate and model the position data of the drone detected by the radar.On this basis,perform data preprocessing on the position data to obtain training data and test data,and the tracking method is trained by the data.Finally,the current statistical model algorithm and interactive multi-model algorithm are introduced to compare the performance.The experimental results show that the improved interactive multimodel algorithm based on the back propagation neural network can better solve the problem of a significant increase in the tracking error that occurs when the target undergoes a step maneuver.On this basis,the support vector regression network is introduced to solve the mismatch between the set and the actual movement of the target.Both improved algorithms can improve the performance of target tracking algorithms and can achieve high-precision tracking of drone in different maneuvering states.