Research On UAV Location Method Based On Machine Learning

In recent years,with the advantages of flexible motion,simple control and small size,UAV has been widely used in civil and military,but it is more difficult to supervise and locate UAV.Therefore,the detection and location of small aircraft in low airspace becomes an important problem to be solved urgently.In the process of UAV target location,the technology of UAV passive location based on mobile communication signal has attracted the attention of many experts and scholars because of its low equipment price and simple location system layout.In the outdoor complex environment,UAV positioning faces the following problems:(1)UAV target volume is too small,resulting in weak reflected signal;(2)UAV moving speed is slow,resulting in the Doppler effect is not obvious;(3)low power of mobile communication base station makes it difficult to separate the signal from the environmental noise according to power;(4)small sampling rate of mobile communication base station leads to large positioning error.Therefore,in this paper,the UAV positioning technology is deeply studied,and the UAV positioning signal separation algorithm based on support vector machine and the UAV positioning error weight compensation algorithm based on random forest are proposed respectively.In order to solve the problem that the small reflector of low altitude slow small target such as UAV leads to weak reflection signal and the low power of mobile communication base station makes it difficult to separate the signal from the environmental noise according to the signal power,this paper takes the separation and extraction of UAV positioning signal as the research target on the basis of machine learning,and focuses on how to extract UAV positioning signal from the environment with serious multipath interference It’s a problem with the number one.This paper proposes a small target location signal separation algorithm based on support vector machine in machine learning.During the training of support vector machine model,the information entropy is obtained by calculating the Euclidean distance between adjacent data sets of small targets,which provides model data for support vector machine mapping high-dimensional space.On this basis,the mapping function threshold soft boundary is added to make the model have the ability of parameter self-adaptive adjustment,which can well adapt to the data differences caused by the flexible movement of UAV.Finally,the observer operation characteristic curve is constructed to obtain the separation result of UAV positioning signal.Aiming at the problems that the Doppler effect is not obvious due to the slow moving speed of UAV and the positioning error is large due to the small sampling rate of mobile communication base station,this paper focuses on the research of UAV positioning error weight compensation technology,and applies machine learning to UAV target positioning.After the target positioning signal is separated,in order to reduce the UAV positioning error,a UAV positioning error weight compensation algorithm based on random forest is proposed.The k-nearest neighbor is used to expand the location data,and cross validation is used to determine the random forest feature parameters and correct the confusion matrix threshold.The classification results of random forest model are used to update the UAV positioning weight matrix,which can effectively compensate the target height data and measure the consistency between the model and the reality.In addition,the calibration UAV is used to estimate the equipment error and Gaussian noise error to correct the positioning results.After theoretical analysis of the proposed algorithm,the superiority of the proposed algorithm is verified by simulation.The simulation results show that the proposed algorithm can effectively separate the UAV positioning signal and noise,and the signal separation accuracy can reach 96.7% under severe multipath interference.The UAV positioning error weight compensation algorithm based on random forest can change the fuzzy area of positioning results into accurate coordinate points,and the positioning accuracy of UAV within 1000 meters in low airspace can reach 2.6 meters.