Oil Slick Thickness Estimation Based On The PolSAR Unsupervised Classification

Using Polarimetric Synthetic Aperture Radar (PolSAR) data to the surface of the oil spill monitoring is one of the new area of Marine remote sensing, full polarization relative literal data, contains rich polarization characteristic information and texture information, and has high efficiency, real-time, not restricted by time, climate and other advantages, so the whole polarization SAR sea surface spilled oil thickness estimation method of the research is of great significance. Compared with the sea ice and other feature information, because the sea wind, waves, and its own chemical reactions, the change of the surface of the oil spill has great dynamic, this adds to the difficulty of the research. On the surface of the oil spill, outline and thickness information is a reflection of it. Design classifier based on the multiple feature fusion strategy, considering the correlation between polarization characteristics, using Mahalanobis distance of fuzzy c-means clustering algorithm is improved, and carries on the oil film thickness gauge. In this paper, the research train of thought mainly includes the following aspects:Firstly, the polarization scattering properties of the surface of the oil spill are analyzed, study and compare the spilled oil can be used for the surface of literal characteristics of oil film thickness estimation.Secondly,this paper uses the improved fuzzy c-means clustering algorithm for estimation of oil film thickness. Fuzzy c-means clustering algorithm, and is the result of k-means clustering algorithm and fuzzy c-means clustering algorithm is given on the number of cluster categories. The improved algorithm is given based on the number of cluster categories,to obtain the initial clustering center automatically. In the pretreatment stage by Mahalanobis distance, calculation of each sample point density, according to the size of the density of sorting, select c a maximum density point as the initial cluster centers.Finally,according to the laboratory to the surface of the oil spill data, as well as the classification research, put forward the strategy of classifier based on feature fusion, used for oil film thickness gauge. According to the feature vector in the accounts for the proportion of different oil film thickness estimation, the distribution characteristics of different weight, carries on the multiple feature fusion; Classifier is designed on the basis of the fuzzy c-means clustering algorithm for the improvement, the algorithm of the pretreatment stage, to join the Mahalanobis distance, automatically calculate the initial clustering center, to carry on the oil film thickness gauge.