Font Size: a A A

Research On UAV Visual Tracking And Control Technology For Aerial Moving Targets

Posted on:2020-05-23Degree:MasterType:Thesis
Country:ChinaCandidate:J R ChenFull Text:PDF
GTID:2392330590472292Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
With the popularity of UAVs and the development of computer vision technology,it has become a hotspot to use visual tracking technology to guide UAVs to track moving aerial targets.Regarding the characteristics of high moving speed and large scale change of aerial,as well as the possible occlusion and interference that may occur during the tracking process,this paper studies and designs a long-term tracking algorithm for aerial targets and a tracking control law of drones in complex scenes,which has been verified perfectly by visual simulation.Firstly,an air target detector based on SVM is designed,and a fast incremental learning algorithm is designed to improve the online learning speed of the detector.The experimental results show that the algorithm improves the training speed by 10.2% and the detection accuracy by 5.3%,compared with the original incremental learning algorithm.Secondly,according to the characteristics of moving objects in the air,this paper designs a longterm target tracking algorithm based on ECO tracker.A scale estimation method based on cubic polynomial is proposed,on which the tracking accuracy is improved obviously in the case of fast scale change.A regularization method for background suppression is proposed to deal with background interference,which focuses on punishing the filter parameters of similar areas between the target and the background.Aiming at the problems of targets' occlusion and loss,aerial targets' detector is utilized to resets the tracking area.This paper validates the algorithm on OTB-100 data set and aerial target data sequence.The results show that the OPE accuracy can reach 0.6432,the average frame rate is 21.2,and the effect is improved obviously.Thirdly,this paper establishes the fixed-wing UAV model and estimates the position of the target in the air.On this basis,the longitudinal and lateral control laws of UAV and the tracking guidance laws based on reference points are designed respectively.The simulation results show that the tracking control law has good performance and can track the maneuvering target stably.Finally,a visual simulation system based on FlightGear is designed.The control program,designed by C++ language,and communication interface between FlightGear and VC++ are completed.The integrated visual simulation is realized successfully.
Keywords/Search Tags:Moving target tracking, SVM incremental learning, ECO tracker, Tracking control law, Visual simulation
PDF Full Text Request
Related items