Based On Polarimetric SAR Image Classification Sub-aperture Decomposition

Synthetic aperture radar has the imaging ability in the all-time and all-weather conditions. Polarimetric synthetic aperture radar is a new kind of SAR system basing on traditional SAR system. By analyzing the ground objects’ backscatter echo signals for the polarimetric radar transmitting electromagnetic waves, rich feature information can be got, then different objects can be recognized. Objects identification and extraction using polarimetric SAR data feature extraction and recognition is one of the important subjects for SAR image interpretation, and is of great practical value and theoretical significance. This research mainly focus on scattering mechanism and image classification method with subaperture analysis.In this paper, research status of the Po1SAR image classification was firstly analyzed, and Po1SAR basic theory also was introduced in detail. Then the commonly used Po1SAR image classification methods were explained, the problems during the feature extraction and classifier selection were pointed in the Po1SAR image classification basing on full aperture image.Secondly, this paper took cloude decomposition for example for introducing polarimetric target decomposition method. After that, target scattering signal characteristics with changing direction was analyzed in the process of Po1SAR imaging. Polarimetric target decomposition characteristics and scattering echo differences of subaperture characteristics were introduced, an approach for classifying the Po1SAR data by integrating polarimetric decomposition, sub-aperture decomposition and decision tree algorithm is proposed herein.Finally, three classification methods are conducted and compared, which are Wishart classification basing on cloude decomposition, decision tree classification basing on cloude decomposition, decision tree classification combining subaperture analysis.The results show that the decision tree model can greatly improve the classification accuracy comparing with Wishart classifier basing on statistical distribution; Besides, the new method proposed in this paper can effectively improve the classification effect compared with classification result of the method basing on full aperture image.