一种结合分水岭与决策树C5.0的极化SAR分类方法

A Supervised Classification Method of Polarimetric Sythetic Aperture Radar Data Using Watershed Segmentation and Decision Tree C5.0

提出了一种利用多种极化特征并结合分水岭算法与决策树C5.0分类器的极化SAR数据分类方法。首先对极化SAR数据进行极化精致Lee滤波,接着对其进行极化分解得到多个极化通道与Pauli RGB图像,改进梯度图生成法并进行形态学分水岭分割与区域合并,最后选择样本构建决策树C5.0分类器并进行分类。实验结果表明,该方法与传统基于像素的分类方法相比精度有显著提高,同时由于使用了较多的极化特征,也使分类精度在一定程度上得到了提高。

A supervised classification method of polarimetric sythetic aperture radar （PoSAR） data using watershed segmentation and Decision Tree C5.0 with many polarimetric channels is proposed. First, the PolSAR data was filtered using the 5 × 5 refined Lee PolSAR speckle filter, and then a Pauli RGB color image and many polarimetric channels were obtained using various algorithms. Then, watershed segmentation on gradient map was made for a homogeneous area and the features of every area were worked out. At last, Decision tree C5.0 was used to deal with the data. The result shows that this method performs better than methods based on pixels, and the classification accuracy is improved with the quantity of polarimetrie characteristic increase.