一种基于SVM的遥感影像分类技术

支持向量机（SVM）应用到高光谱图像分类中有较好的识别效果。但用它来分类数据量大、维数高的高光谱图像时，就会遇到如何选择最佳惩罚因子和最优权向量系数的问题。提出一种改进的多类支持向量机分类方法，在OAO-SVM分类结果的基础上进行二次分类，以改善错分样本较多的类别之间的混淆程度。实验表明，二次分类的多类支持向量机方法是有效的。

A Remote-sense Image Classification Technology Based on SVM

The application of SVM in high spectral image classification is of fairly good recognition effect. However, the problem of how to select the best penalty factor and weight vector coefficient wouldappearinusing it for the classification of large-volume and high-dimension high spectral image. Thus amodified classification method of multi-class SVM is proposed, and a secondary classification based on OAO-SVU classification is implemented, thus improving the mixing degree between among the classes with multiple erroneously-classified samples. The experiment proves the effectiveness of the proposed method.