PolSAR数据区域海冰LSSVM分类

LSSVM Classification of Sea Ice for Segments of PolSAR Data

本文给出了一种结合影像分割和最小二乘支持向量机(Least Square Support Vector Machine,LSSVM)的海冰监督分类方法。考虑到SAR数据的斑点噪声影响分类精度,首先利用分水岭算法和等级区域合并执行影像分割,利用分割后的区域代替像素开展分类研究,由于分割尺度影响分类精度,文中利用LSSVM算法对不同尺度下的分割结果执行分类,通过评估分类精度确定最优分类结果及其对应最优分割尺度,从而克服经验选择分割尺度缺乏理论依据的不足。利用两组SIR-C数据验证了算法,实验结果表明本文算法的总体分类精度超过85%,能较好的识别不同类型的海冰。

A method of sea ice classification is presented in this paper with combination of image segmentation and LSSVM supervised classification.Because of classification accuracy of SAR data affected by speckle noise,the image is segmented initially by watershed algorithm and then merged step by step to retrieve final segments called regions which instead of pixels are used to classification.The final classification accuracy depends on the segmentation level,so different level segmentation is used to classification by LSSVM.The optimum segment level is selected corresponding to the high classification,rather than by experience choice,while the optimum classification is realized.Two set of SIR-C data is used to validate the method,the result shows that the overall classification accuracy exceeds 85%,proved the type of different sea ice can be distinguish by this method.