基于支持向量机和Getis因子的高分辨率遥感图像分类

Study on Classification of High Spatial Resolution Remotely Sensed Imagery with SVM and Local Spatial Statistics Getis-Ord Gi

采用支持向量机对具有RGB 3个波段、分辨率为0.32 m的航空摄影图像进行实验,首次根据表示空间聚集程度的局部Getis因子完成分类。结果表明:1)当应用基于线性、多项式、径向基和Sigmoid 4种常用核函数的SVM进行分类时,基于径向基的SVM分类精度最高,总体精度超过91%。2)从原始图像计算出局部Getis因子,该指标可用于图像分类,且分类精度与局部Getis因子的步长有关;在步长小于变异函数变程的条件下,应用径向基SVM的总体分类精度达95.66%,高于直接使用原始图像RGB波段光谱信息的分类精度,因此局部Getis因子在高空间分辨率遥感图像分类中具有应用和研究价值。

Land classification for high spatial resolution remote sensing images is an important topic in many applications. In this paper, the support vector machine （SVM） algorithm was utilized to tackle the classification of a 3-- band image from airborne digital sensor system with ground resolution of 0. 32 meters. Firstly, the original image was classified using SVM of four common types of kernel functions, namely linear, polynomial, RBF and sigmoid function, and the SVM with RBF kernel function can achieve the most satisfactory result with statistical overall accuracy over 91%. On the other hand, Getis--Ord Gi, one type of local spatial statistics to determine clusters of similar values, had been calculated based on the original spectral image with varying lags from 1 to 10. Classifying G with lag of 3 other than the original spectral image,an overall accuracy of 95. 66% was achieved using SVM based on the RBF kernel function. The result of the experiment shows that Gi with lags less than the variogram range can substitute for the original spectral image to improve the classification accuracy between features with similar spectral characteristics like trees and lawns, as a result, to increase the overall classification accuracy.