Continuous Information Reconstruction Of Atmospheric Polarization Pattern Based On Neural Network

Atmospheric polarization pattern as earth's natural attribute,which contain rich information of the distribution of the optical properties and stability of the atmospheric polarization pattern in the sky,even under the complex atmospheric environment factors,the atmospheric polarization pattern still presents a series of continuous distribution of time and space,therefore in the autonomous navigation,target detection and other fields has a broad application prospect.But in the practical polarization data acquisition still has certain problem,in order to solve the atmospheric polarization pattern measured experiment can appear sometimes data missing or short data sets,this thesis designed a kind of local atmospheric polarization modes based on sequence information network information reconstruction,and validated by simulation and experiment of the polarization data in this thesis,the feasibility and the robustness of the proposed method;Secondly according to the measured experiment sometimes local data missing and more difficult to obtain the measured polarization data under different conditions of the problem,this thesis designed a based on limited samples context information network of atmospheric polarization pattern information reconstruction on the simulation data and the measured data,and compared with other new methods,Experimental results demonstrate the superiority and robustness of the proposed method.The main research work is as follows:(1)Through the study of the basic theory of atmospheric polarization mode and the distribution characteristics of atmospheric polarization mode under different weather conditions,combined with the outstanding characteristics of atmospheric polarization mode found by previous scholars,such as stability and testability,on this basis,the problems in data set acquisition in the measured experiment were analyzed.Compare the existing techniques and analyze why the problem of missing data in the data set and difficult access to data under different conditions cannot be solved.(2)According to the atmospheric polarization pattern measurement experiment because equipment or weather and other reasons,caused by the measured data set data missing or data discontinuity problems,highlight the characteristics of atmospheric polarization mode and GAN research of neural network algorithm,this thesis build a refactoring atmospheric polarization modes based on sequence information information network,The results show that the local atmospheric polarization mode information reconstruction network designed in this thesis is feasible and robust to some extent.(3)Due to the limitation of the physical characteristics of atmospheric polarization information acquisition device and the occlusion of foreign bodies in the sky sometimes occurring in the measured experiment,the global atmospheric polarization mode data can be obtained by this method.In addition,it is difficult to obtain the polarization data under different conditions in the measured experiment.In the third chapter of this article for further research and combining the Few-shot learning algorithm,according to the above problem,this thesis designed a kind of context information of the atmospheric polarization pattern based on limited sample refactoring network,information on the simulation and measured data reconstruction experiments,and compared with other information reconstruction algorithm,the experimental results verified the superiority and the robustness of this method.