Study On Monopulse Forward-looking Imaging And Classification Recognition In Polarization Radar

Polarization radar introduces polarization information,which enriches the knowledge of the target electromagnetic scattering characteristics and provides a way for comprehensive interpretation of target information.Compared with the traditional single-polarizatiom radar,the polarization radar has the information advantage of the multi-polarization channel,and the performance advantage of the complete polarization domain information joint processing.Starting from monopulse forward-looking imaging,this paper covers the imaging of single-target,dual-target and group targets in a radar beam,focusing on the advantages of polarization radar for monopulse angle measurement and polarization information for target recognition.Finally,the role of polarization color information in the terrain classification and recognition of features is studied.Firstly,the monopulse forward-looking imaging with one single target within a radar beam is simulated.Then a noval monopulse forward-looking imaging algorithm based on Levenberg-Marquardt(LM)optimization is proposed to solve the problem that the two unresolved in the beam.The core idea of the algorithm is to solve the maximum likelihood estimation problem based on LM optimization to obtain Direction of Arrival(DOA)estimation of two unresovled targets.The simulation results show that two targets within a forward-looking radar beam can be resolved and relocated accurately utilizing the proposed algorithm.Meanwhile,the comparison with other algorithms shows it has higher DOA estimation accuracy,less computational complexity and a wider range of angle interval adaptability.For the monopulse forward-looking imaging in polarization sensitive array radar,the relevant information between polarization channels is fully utilized,and the polarization monopulse angle measurement is realized by virtual polarization matching.The angle measurement accuracy is better than that of single-polarization radar,and it is unaffected by the wave polarization.After expanding the target into group targets,a polarimetric monopulse forward-looking imaging algorithm based on Bayesian Adaptive Direct Search(BADS)is proposed,the nonlinear optimization is solved by BADS from the maximum likelihood estimation problem.Then the DOA and polarization phase descriptors of the unresolved targets are estimated.Combined with the principle of polarization measurement,we can obtain the polarization scattering matrix.The results show that the estimation accuracy of the DOA and polarization phase descriptors is greatly improved compared with the gradient descent method.The imaging results can reflect the position of the target nicely,and the algorithm can accurately identify three targets in the group under high SNR conditions,including a dihedral angle and two trihedral targets.For the terrain classification of Polarimetric Synthetic Aperture Radar(PolSAR),a new method for terrain classification combined with color information in polarimetric SAR image is proposed,because the polarization target decomposition provides the color representation.The image generates superpixels,and the superpixels are classified as a whole.It not only improves the classification accuracy,but also achieves excellent noise immunity.Finally,the terrain classification accuracy can be further improved by the Complex-Valued Convolutional Neural Network(CV-CNN)classifier which considering the phase information of the polarimetric SAR.