基于支持向量机的矿区土地覆盖变化检测

Land cover change detection over mining areas based on support vector machine

利用支持向量机(SVM)对两类问题良好的分离性能,将其应用于矿区土地覆盖变化检测,实现了基于SVM的变化检测算法.该算法计算了多时相遥感数据的差值影像,利用SVM将全部像素分类标记为变化和不变化两个类别,在不变区域中选择训练样本,对变化区域进行分类,获得前后时相地物类别信息,构建变化转移矩阵,描述详细变化信息.应用多时相先进对地观测卫星(ALOS)遥感数据对矿区土地覆盖变化进行试验,并与变化矢量分析、差值阈值法进行对比,结果表明:基于SVM的变化检测方法具有更好的检测效果,能够提供全面的变化类别和方向信息,可以有效应用于矿区土地覆盖动态监测.

A supervised change detection approach based on support vector machine （SVM） is proposed by making full use of the SVMrs good capacity for two-class separation. The technical flow of change detection based on the SVM is designed and implemented, in which the differ- ence images are generated from multi-temporal remote sensing images at first, and then all pix- els are labeled as changed and unchanged by binary SVM detector. In order to obtain the de- tailed land cover change information, samples are selected from the unchanged areas to train a SVM classifier, by which land cover of change areas are classified, and a change matrix is con- structed. Multi-temporal advanced land observing satellite （ALOS） images over a mining area are used as experimental data. By comparing the proposed SVM-based method with change vec- tor analysis and image differencing, it is concluded that the SVM-based method can obtain higher accuracy than other methods and provide detailed change type and direction information at the same time. The proposed SVM-based change detection method is a promising approach for land use/cover change monitoring over mining areas and other regions.