Classification Of Polarimetric SAR Image Based On Polarization Ratio Feature And Affinity Propagation Clustering

Compared with synthetic aperture radar(SAR),Polarization Synthetic Aperture Radar(Pol SAR)data contains more abundant information of terrain and target.Moreover,due to its all-day and all-weather capabilities,remote sensing image processing technology based on polarimetric SAR data has been rapidly developed and applied.Polarimetric SAR image classification is one of the most important applications.Supported by the National Natural Science Foundation of China(Pol SAR image classification based on co-training and sparse representation,No.61173092),and the National Natural Science Foundation of China(Pol SAR image classification based on generative confrontation network,No.61771379),This paper propose three improved unsupervised Pol SAR image classification methods,which combine polarization characteristic and affinity propagation clustering algorithm.Main tasks are as follows:We defined a polarization ratio feature.This method put forward a polarization ratio feature by analyzing different channel data of the polarization scattering matrix.And combine this feature with SPAN and scattering power entropy to finish the initial segmentation of Pol SAR image.Then,we use the Wishart test statistic to perform an agglomerative hierarchical clustering to obtain the segmentation results with different numbers of clusters.This method can further improve the accuracy of image classification.We presented an improved affinity propagation clustering algorithm based on distance information and density information.The traditional AP algorithm only takes into account the distance information of data when calculating the similarity between sample points,ignoring the influence of density information on similarity measurement.To solve this problem,this paper introduces the distance and density information in the data to improve the similarity matrix in the AP algorithm,and make it better describe the actual relationship between sample points.Then,We use the improved AP algorithm to achieve data clustering adaptively.Compared with the current classical Pol SAR image classification methods,this method has better classification accuracy.An affinity propagation clustering algorithm based on improved preference is proposed and applied to the classification of polarimetric SAR images.In general,AP clustering algorithm assumes that each sample become the candidate representative point with equal probability.So,the preferences are same in the similarity matrix.In the actual case,this assumption is not reasonable.In view of this situation,this paper has made a corresponding improvement,so that the probability of becoming the representative point for each sample is associated with the similarity between it and the other sample points.The greater the similarity,the greater the preference.The probability that the sample point will be selected as the representative point is higher.Then,we use the improved preferences to update the similarity matrix,and apply the matrix to the AP algorithm to realize the final clustering result of image.The classification results of four groups of polarimetric SAR data show that this method has high classification accuracy and strong universality.