Research On Target Tracking Technology Based On Multi-Rotor UAV

With the development of the new military revolution and the rapid expansion of the UAV market,the development of UAV target tracking technology has been closely linked with the interests of national security and the work and life of citizens.As a popular research field in the computer field,target tracking has always been a hot research topic both at home and abroad.This thesis studies the intelligent target tracking algorithm.Two kinds of target tracking algorithms,MF-KF-Camshift and Staple-l,are proposed,which are proved by the multi-rotor UAV platform the effectiveness of the algorithm.Specific work includes the following aspects:Firstly,Proposed MF-KF-Camshift algorithm.Based on Camshift,this algorithm adopts multi-feature fusion method to solve the problem of performance degradation in complex background based on color features.At the same time,the Kalman filter algorithm is introduced to solve the problem of insufficient robustness to blockade.In addition,the self-built video library verifies the performance of the algorithm.Experiments show that the algorithm overcomes the difficulty of UAV target tracking based on real-time guarantee.Compared with the original algorithm and the popular Kalman filter improvement algorithm,The proposed algorithm respectively improves the accuracy of about 30.5% and 11.8%,enhances the robustness and realizes the target tracking of the UAV.Secondly,Aiming at the template drift problem of tracking algorithm in the tracking process,this thesis proposes an evaluation index of tracking confidence based on Staple algorithm,and evaluates the advantages and disadvantages of tracking results and the optimization of updating strategy.By introducing the re-detection module,Re-test the target.Combining with the evaluation index of tracking confidence and the re-detection module,a Staple-l algorithm which can be tracked for a long time is proposed.The experiment of VOT database shows that the proposed Staple_l algorithm improves the robustness by about 11.5% compared with the original algorithm while ensuring the accuracy.Thirdly,With Raspberry Pi processor as the image processing equipment,a multi-rotor UAV platform based on open-source flight-controlled APM was built by using aerial camera.In addition,the application of MF-KF-Camshift and Staple_l proposed in this thesis is realized on the constructed platform.The feasibility of the proposed algorithm is verified by real-flight experiments.The work of this thesis is practical.The MF-KF-Camshift algorithm and Staple_l algorithm are proposed to realize the target tracking of UAV and solve the problem of lack of tracking performance of multi-rotor UAV platform.The target tracking applied to the UAV platform provides the foundation for the subsequent realization of fully autonomous intelligent UAV,with high practical significance.