融合多光谱与SAR影像的地物分类研究

Study on Classification of Ground Objects with Multispectral and SAR Images

为提高基于遥感影像的地物分类精度,以河北省遵化市为研究区,利用Landsat-8 OLI多光谱影像和Sentinel-1A SAR影像开展基于影像融合的地物分类研究。首先,结合波段相关性分析与最佳指数因子确定Landsat-8 OLI影像中参与地物分类的最佳波段组合,基于灰度共生矩阵提取Sentinel-1A影像中参与地物分类的最佳纹理特征;然后,采用Gram-Schmidt方法融合Landsat-8 OLI的最佳波段组合与Sentinel-1A的最佳纹理特征;最后,利用支持向量机分类法对融合影像进行分类,并对分类结果进行比较分析。结果表明:基于融合影像的地物总体分类精度达87.30%,较单一的Landsat-8OLI影像提高15.28%,对水体和建设用地的区分程度有一定提高。

It is very important to obtain land use information quickly and accurately for proper utilization and development of land resources. In this paper,Landsat-8 OLI and Sentinel-1 A images are utilized to classify land use in Zunhua City,Hebei Province. The best band combination of Landsat-8 OLI image is selected according to the correlation analysis and the Optimum Index Factor,and the best texture features of object is extracted based on gray level co-occurrence matrix in Sentinel-1 A image. Then Gram-Schmidt method is applied to fuse the best band combination of Landsat-8 OLI with the best texture features of Sentinel-1 A. Finally,the classification of the fusion image is performed using the support vector machine method,and the classification results are compared and analyzed. The results show that the accuracy of ground feature classification based on fused images reached 87.30%. Compared to single Landsat-8 OLI image classification,the accuracy increased by 15.28%,there is a certain increase in the degree of distinction between water bodies and residential land. This study provides a theoretical basis for fusion of multi-source remote sensing images based on their effective information to improve the classification accuracy.