结合纹理信息的极化SAR影像分类研究

Polarimetric SAR Images Classification Research Combined with Texture Information

结合Gabor小波、灰度共生矩阵和Fast ICA方法提取的纹理信息,利用支持向量机分类器对单极化SAR影像进行分类研究。首先利用精致Lee滤波器对影像进行去噪处理;然后采用灰度共生矩阵和Gabor小波提取影像纹理特征,利用Fast ICA算法对纹理特征进行降维分析;最后将降维后的纹理特征与强度特征结合,采用支持向量机分类器进行分类;采用北京地区Terra SAR-X影像对该方法进行实验,结果表明,纹理信息的引入使极化SAR影像分类精度得到提高。

Combined with Gabor wavelet, gray level co-occurrence matrix(GLCM) and FastI CA method, we extracted the texture information and used support vector machine classifier to classify the single polarization SAR images. Firstly, we used enhanced Lee filter to deal with the noise of the images. And then, we used GLCM and Gabor wavelet to extract image texture features, and used FastI CA algorithm to reduce the texture feature dimension. Finally, combined with texture feature and strength feature, we used support vector machine classifier to classify. We tested the method by the TerraS AR-X satellite images in Beijing area. The results show that the introduction of the texture information improves the accuracy of polarimetric SAR images classification.