Resolution Enhancement For ISAR Image Based On Joint Dictionary Learning

Recently,machine learning and sparse representation has been applied in various fields,especially in the field of signal processing.An ISAR imaging sparse reconstruction algorithm based on joint dictionary learning is proposed based on the similarity between optical imaging and ISAR imaging.The main work of this paper is as follows.The first chapter introduces the research status of image super-resolution and summarizes several classical methods and working principles of ISAR resolution enhancement technology.The second chapter analyzes the imaging principle of ISAR and the principle of sparse imaging.The turntable imaging model under small angle is introduced.Signals can be denoised by sparse representation via dictionaries,and ISAR imaging has the advantage of not requiring pre-estimation of scattering points.So we introduced the application of sparse representation in ISAR imaging.In third chapter,a sparse reconstruction model is proposed,which is based on the joint dictionary learning method of low-resolution and high-resolution image patches to enhance the image resolution.In ISAR imaging,the echo is equivalent to a similar scattering center model when an object is illuminated by a radar signal with the same center frequency but different bandwidth.So it is reasonable that the low-resolution ISAR image of the object shares the same sparse representation coefficients with its high-resolution ISAR image.When low-resolution ISAR images are sparsely represented by low-resolution dictionaries,high-resolution ISAR images can be reconstructed based on high-resolution dictionaries because of similar sparse representation coefficients.The fourth chapter mainly studies the use of complex radar images to enhance image resolution.In terms of dictionary optimization algorithm,this paper proposes a dictionary training algorithm based on RBM network model.Hidden layer nodes in the model represent sparse representation,and explicit layer nodes represent high/low resolution image data.Therefore,the weight in the model represents the dictionary.Finally,the simulation results show that the dictionary learning algorithm based on RBM network is the best algorithm for sparse restoration.The complex radar image data is used for training and testing in this chapter.The real and imaginary parts of radar data are separated.The final experimental results show that the simulation of complex radar data has better effect.