Font Size: a A A

Telescope State Measurement Based On Artificial Intelligence Chip

Posted on:2021-12-12Degree:MasterType:Thesis
Country:ChinaCandidate:X Y LiFull Text:PDF
GTID:2480306110496974Subject:IC Engineering
Abstract/Summary:PDF Full Text Request
The telescope does not only play an important role in the astronomical observation,but can also be used to observe space targets and space debris.It is also a basic instrument for high energy laser systems.The impulse response of the optical system,known as the point spread function(PSF)can describe the state of the telescope.In order to realize the intelligent control of the telescope,this paper studies a method based on the PSF of the telescope,using an artificial intelligence chip to measure the state of the telescope in real-time.As a data-based PSF modelling method,the principal component analysis method(PCA)has been widely used in PSF modelling of space-based telescopes and ground-based telescopes.However,since PCA requires large amount of data for statistical analysis,it does not meet the real-time requirement.The PCA method is more suitable for data post-processing than for real-time intelligent control of telescopes.The denoising autoencoder(DAE)is based on the basic autoencoder(AE),adding constraints to make sure it can learn the manifold distribution characteristics of data.The trained DAE can reconstruct the original data from the destroyed data.DAE is trained by PSF with different noise levels.After training,the DAE can map any images of celestial objects obtained by the wide-field telescope directly to the original PSF.Furthermore,we build a convolutional neural network(CNN)to learn the map between PSFs and the telescope states.Therefore,the artificial intelligence chip equipped with the neural network can realize the online measurement of the telescope state with telescope star images.We find that the method proposed in this paper is robust by testing simulated data and real data.We further install the above model on the Nvidia Jetson Xavier,and find that the data processing speed meets our requirements,which preliminarily demonstrates its ability to adjust the state of telescopes,providing a method in the development of intelligent astronomical optical instruments.
Keywords/Search Tags:Point Spread Function, Denoising Autoencoder, Artificial Intelligence Chip, Telescope
PDF Full Text Request
Related items