基于SVM的遥感影像土地利用变化检测方法

Land Use Change Detection Based on SVM

利用遥感影像的直方图统计量作为基本检测特征,采用支持向量机(Support Vector Machine,SVM)理论对淄博某地区的土地利用变化情况进行检测。为验证SVM方法的检测精度,分别使用Fisher判别准则、类别可分离判据等方法进行对比分析。遥感数据来自两个时相分辨率为0.6 m的Quickbird影像,相对于传统TM、SPOT影像,该方法能够提取出小范围内的土地利用变化情况,适合区县级范围内的土地变更调查。

This article researched on land use change detection through sub-meter resolution images. Several well-known methods were employed here to validate detection＇ s results, such as theory of Fisher transformation, the primary component analysis （PCA）, the method of nearest distance and the method of support vector machine （SVM）. In order to increase detection＇ s precision, 12 features from histogram were brought up for change detection. Further more, this article used Quick-bird＇ s images （0.6 meter resolution） , which can obtain tiny change information comparing with the images.