Research On Multi-source Information Based Guidance Technology For Unmanned Aerial Vehicle Autonomous Carrier-landing

As carrier landing is a curial part among the combat chain of the carrier based Unmanned Aircraft Vehicle(UAV),the research on the guidance technology for UAV automatic carrier landing is significant.Multi-source information fusion is a promising method for UAV automatic carrier landing,and this paper focuses on the guidance technology for UAV autonomous carrier-Landing.Firstly,the advantages,disadvantages,error and the information flow of the precision guidance radar,Global Position System(GPS)and machine vision are systematically compared in this paper.An information fusion scheme for UAV automatic carrier landing is proposed on the basis of the comparison,and the scheme contains three parts: enhancement of vision information,automatic recognition of carrier and relative pose estimation of the carrier and UAV.Secondly,a multi-source image fusion based vision information enhancement method is proposed.The K-means Singular Value Decomposition(K-SVD)was used to train an over-complete dictionary from the training samples.The sparse coefficients of the visible images and the infrared images are computed on the basis of the over-complete dictionary.The sparse coefficients are fused by the fusion rule,and the fused images are finally reconstructed.The experiments demonstrate that the method is effective for the enhancement of the vision system.Thirdly,a multi-feature fusion based carrier automatic recognition method is proposed.The Spectral Residual(SR)is used for candidate regions extraction of the carrier.The Scale-invariant Feature Transform(SIFT)descriptors are extracted from the image,the higher level sparse features are obtained by sparse coding.The global and local features are fused by the structure of Spatial Pyramid Matching(SPM).The carrier is recognized by a linear Support Vector Machine(SVM).The experiments demonstrate that the fusion of multi-feature increase the robustness of the recognition,and the carrier can be recognized under different kinds of weather.Finally,the Inertial Navigation System(INS)and vision fusion based relative pose estimation method is proposed.The initial value of the pose information is proposed by INS.The pose information is further optimized by Orthogonal Iteration(OI),and the optimized pose information is fed back to INS.The experiments demonstrate that the fusion algorithm solve the jump of the estimation of vision algorithms,and greatly decreased the error.