基于多核SVM的高光谱影像植被精细分类

Sophisticated Vegetation Classification Based on Multiple Kernel SVM Using Hyperspectral Images

植被自身复杂的生长环境和垂直分布结构，使得高光谱影像中的植被特征受到大量异构信息影响。在对植被进行精细分类时，随着植被类别的增加，植被样本信息量大大增加，但植被各类别之间的可分性却在下降，采用单核映射方式对所有植被样本进行处理的分类方法难以得到可靠的分类精度。多核学习方法能够以全新的核函数映射方式对复杂的样本信息进行处理，本文将多核学习方法引入植被精细分类中，提出基于多核SVM的高光谱影像植被精细分类方法，实验结果表明该方法可以显著提高分类精度，在树种识别、精细农业等方面具有广泛的应用前景。

Due to the complicated growth environment and vertical distribution structure, the vegetation characteristics in hyperspeetral images are influenced by a large amount of heterogeneous information. During the sophisticated classification of vegetation, the amount of vegetation sample information increases greatly with the enrichment of vegetation types, but the separa- bility between different classification decreases. Therefore, it is difficult to achieve reliable classification accuracy by processing all samples with classification methods based on single kernel function. By contrast, muhiple kernel learning method using new kernel function mapping mode can deal with complicated sample information. The new method is introduced to solve vegetation classification problems, and sophisticated classification method based on multiple kernel SVM is proposed. Sophisticated vegeta- tion classification experiment is conducted using hyperspectral images with real ground vegetation data, and the results show that the method can significantly improve the classification accuracy and has a wide application in species identification, precision agriculture and other fields.