Research On Fast Radio Burst Localization Based On The Focal Field Features

Fast radio burst(FRB)is one of the main research objects of time-domain astronomy.Because FRB events are usually characterized by short duration and extremely high explosive energy,so it is considered to be an important probe to study cosmology.However,the kind of sudden astronomical phenomena are difficult to predict in space and time.This poses a difficult problem for the transient airspace detection capability of radio telescopes,which is the main observation tools of radio astronomy.Based on the research background of FRB localization,this paper aims at expanding the field of view of radio telescope and improving the localization accuracy of radio targets,combined with deep learning technology,proposes FRB localization method based on the characteristics of the focal surface field.The content of the research is as follow:(1)Carry out research on the correlation between the focal field distribution geometry,texture characteristics of the focal plane of the radio telescope and the azimuth of the radio target.The professional electromagnetic simulation software FEKO is used to simulate and analyze the electromagnetic field distribution generated by radio target in each spatial direction on the focal plane.The simulation results show that the geometric characteristics of the focal field distribution are related to the orientation of the radio target,thereby establishing a mapping relationship between the direction information of the radio target and the focal field distribution characteristics.(2)Aiming at the structural characteristics of the FAST telescope feed system,a sampling method of the longitudinally moving feed system is proposed,which can obtain the three-dimensional electromagnetic field distribution,and combined with deep learning technology,a FRB location algorithm based on the attention mechanism is designed.Due to the large electrical characteristics of the horn cluster feed,three-dimensional field distribution we have obtained is very sparse.First,in order to obtain a complete three-dimensional field distribution information,the super-resolution reconstruction of the sparse three-dimensional field distribution is performed.Then,based on the reconstruction results,FRB location is performed.At the same time,attention mechanism and perceptual loss are introduced into the algorithm to improve the reconstruction effect and localization accuracy.The simulation experiment results show that the three-dimensional electromagnetic field distribution under the limit of the number of feeds which can be accurately reconstructed by the proposed algorithm,so that the radio targets inside and outside of the radio telescope can achieve high localization accuracy,and to a certain extent,the transient field of view of the radio telescope is expanded.(3)In view of the structural characteristics of the FAST feed system,the sampling method of the rotating feed system is proposed,which increase the sampling number of the focal field,and at the same time a FRB location algorithm based on LSTM-VAE is designed.The structure of the algorithm is mainly based on VAE.Meanwhile,in order to dig the intrinsic correlation between different sampling sequences,the encoder and decoder are implemented by LSTM neural network,combined with MLP to achieve FRB localization.The simulation experiment results show that the algorithm can effectively estimate the radio targets inside and outside the field of view,and achieve a localization accuracy much higher than the telescope angular resolution.