积雪面积是融雪径流模型中变量数据输入之一,准确的获取雪盖范围是进行流域尺度融雪水文过程研究的关键,在水资源管理及洪水预报中具有重要意义。本文以天山山区中段为例,利用MODIS数据,提出了结合混合光谱分解的积雪分量及灰度共生矩阵提取的纹理特征的SVM分类方法,对研究区积雪面积信息提取进行了研究。结果表明:通过利用混合光谱分解的积雪分量作为SVM的特征输入,总体分类精度比传统SVM分类结果有了一些提高。同时考虑结合基于灰度共生矩阵提取的纹理特征用于分类中,总体精度比传统SVM方法提高了1.081%,制图精度达到了99.01%。本文提出的分类方法能够适应特征组合之间的非线性关系,从而能提供更多的区域地物空间分布信息,能够调整无样本地表类型地区的积雪面积反演,对今后的融雪水文过程研究有重要意义。 The snow area is one of the variable data input in the snowmelt runoff model. Accurately obtaining the snow cover is the key to study the hydrology and snowmelt flow in the basin scale, which is of great significance in water resources management and flood forecasting. Taking the middle section of Tianshan Mountains as an example, this paper presents a SVM classification method based on MODIS data, which integrates mixed snow spectral components and texture features extracted from grayscale co-occurrence matrix, and studies snow area information extraction in the study area. The results show that the overall classification accuracy is improved more than the traditional SVM classification by using snow spectral components decomposed by mixed spectra as the feature input of SVM. At the same time, the texture features extracted based on the gray level co-occurrence matrix are considered for classification. The overall accuracy is improved by 1.081% compared with the traditional SVM method, and the accuracy of the drawing is up to 99.01%. The classification method proposed in this paper can adapt to the non-linear relationship between feature combinations and thus provide more spatial distribution information of regional geomorphology, and can adjust the snow surface area inversion without sample surface types. The study on the future snowmelt hydrological process There’s important meaning.