Polarimetric SAR Image Classification Based On Intensity Statistics Sparse

Polarimetric Synthetic Aperture Rader(POLSAR) is a multi-parameter and multichannel Radar Imaging System, which sends and receives different polarization radar imaging system by electromagnetic wave, get full polarization SAR image features of all scattering information on the four channels, namely the emission levels receive channels(HH), horizontal vertical launch receiving channel(HV), vertical launch levels receive channel(VH), vertical launch vertical receive channel(VV), four channel includes the full polarimetric scattering information in every resolution unit on the ground. And it is also the biggest advantage of the full polarization SAR image relative with literal SAR image. In polarization SAR image classification problems, how to extract classification characteristics from polarization SAR scattering information; how to apply all known polarization scattering information become the key for success or failure and quality problem of classification. The full polarimetric SAR image provides four channel polarization scattering information. For how to make good use of the polarization scattering information as well as the connection between the four scattering channel, this paper puts forward the assertion and methods applied to the full polarimetric SAR image classification. The main contents are as follows:1. We proposed a new thesis- the polarimetric scattering intensity in the polarimetric scattering intensity matrix of the full polarimetric SAR has the similarity on four channels. Polarization coherent matrix and the method for extract features from polarization coherent matrix used for polarization SAR image classification has been very mature. While the polarization coherent matrix scattering contains all scattering information of terrain scattering target, but it blurred this feature of the polarimetric SAR four channel. In order to overcome this defect, this paper discussed the advantages of the polarization SAR of scattering intensity matrix and coherent matrix, as well as the sparsity of scattering intensity into the polarization SAR scattering intensity matrix in the four channels. For feature extraction in the polarization SAR image, these would be a major breakthrough.2. We proposed a fine classification method---based on full polarization SAR on each channel scattering intensity statistics sparse polarization SAR image classification and Classification of dictionary migration. To the front of the similarity of the polarimetric scattering intensity in the polarimetric SAR between channels, for a specific feature of target scattering intensity on the three channels, we make statistics. By KSVD characteristics of dictionary learning algorithm for a particular object dictionary; through orthogonal matching algorithm(OMP) won the sparse characteristics of the polarization scattering intensity on the three channels. Use these sparse features which contain three channels relations of complete polarization SAR image classification. At the same time it can also be moved the dictionary of the specific goal into unknown polarization SAR image data, finally implement specific target classification.3. We proposed another new method – in the full polarization SAR image, scattering intensity on each channel statistics sparse method of image classification and recognition under the tensor model. By tensor descript Polarization scattering intensity statistics sparse, they more can reflect the full polarization SAR structural information between the three channels, and enrich the classification feature. What's more, they better realize the polarization SAR image classification, at the same time they can also put the specific goal after tensor representation dictionary migrated to unknown polarization SAR image data, achieved the identification of specific goals.