高分辨率遥感卫星影像在土地利用分类及其变化监测的应用研究

研究了IKNOS米级高分辨率遥感影像在大比例尺土地利用图件更新中的应用技术,提出采用基于知识的土地利用覆盖分类以及变化监测系统方法,首先利用NDVI植被指数和半方差纹理特征的知识进行影像大类区域分割;其次结合光谱知识对各影像区域进行详细分类,同时利用区域生长技术与地类空间知识进行区域分类;第三步是分类后处理与变化信息提取,利用基础图件提供的知识与各区域分类进行比较以发现变化的区域。北京房山良乡试验区的试验表明,Kappa系数为0.912,总精度为0.938;变化信息错误率为13.69%,基于知识的分类与变化信息自动提取可以为在GIS/RS环境下的目视数字化提供目标,加速土地利用基础图件的更新作业过程

Application of High-Spatial IKNOS Remote Sensing Images in Land Use Classification and Change Monitoring

With the merge of the meter based high spatial remote sensing satellite, the sources for updating of the large scale land use base maps were provided. The practical technique to update the large scale land use base maps with the meter based high spatial remote sensing images was studies. The method of the land use/land cover classification and change information extraction system based on knowledge is used. First, the NDVI and semi variogram texture characteristics are used to segment the building area, vegetation area, bare land and water area, grass & forest land, road area, the results act as hierachy controller; Second, the spectrum, vegetation indexes and texture characteristics knowledge are applied to classify these regions specifically. At the same time, the region growth technique and spatial entity knowledge are used to modify the classification results; Third, the comparison of land use base map and the remote sensing classification can identify the change regions and correct the classification errors. The experiment results of the test region in Fangshan county of Beijing demonstrate the accuracy of the classification and change information extraction are relatively high. The Kappa coefficient is 0.912,the overall accuracy is 0.938 and change information error is 13.69%. The visual digital target can be supplied through the classification and change information extraction based on knowledge. This research can help reduce the work task and accelerate the visual screen updating process. So it will be widely applied in updating the land use base maps during the survey of land resources.