太湖湖滨敏感区的土地利用遥感分类研究

Land Use Classification based on Remote Sensing Image in Taihu Lake Lakeside Sensitive Area

近年来太湖流域水体污染日趋严重,土地利用是重要的环境变化影响因子,对太湖湖滨敏感区土地利用分类研究具有重要意义。研究基于2010年ALOS多光谱遥感影像,以太湖流域上游的武进港、直湖港流域为研究区,根据研究区实际状况和研究目的,建立太湖流域上游湖滨敏感区的土地利用/土地覆被分类系统,并用于该地区的面向对象遥感分类,研究通过影像的多尺度分割,获得不同层次的影像对象,在不同层次设置对应的分类规则,以充分利用影像中地物的光谱、纹理和不同层对象相互关系等信息,从而提高分类效果。研究表明:在面向对象多尺度影像分割的基础上,基于决策树建立多个分类规则的分类方法,能够有效提取建设用地、道路、水体等几类信息,分类总体精度达到88.00%;同时,该地区主要土地利用类型如耕地、农村居民点和城镇居民点的分类精度也较高,这也表明该分类方法对整个太湖流域以及其他平原河网地区的土地利用相关研究具有一定的实用价值。

In recent years, water pollution has made great crisis in Taihu Lake Watershed. Land use is seemed as an important factor for the environmental changes,so the land use classification study of Taihu Lake lakeside sensitive area has great significance. According to the local circumstance of land use and the research objectives in the study area,a novel land use/land cover classification schema is established for the Wujingang River ＆ Zhihugang River Watershed in the upstream areas of Taihu Lake Watershed,then the reliable watershed land cover/land use information can be acquired by the means of images classification based on medium-resolution remote sensing images. In this study, ALOS multi-spectral remote sensing ima- ges in 2010 are used for the data sources, object-oriented image classification are performed after the multi- resolution image segmentation, then several classification rules based on decision tree methods are built to effectively extract the land use/ land cover information such an construction lands, road and water body. The classification results show that based on such an object-oriented image classification method has a higher accuracy,and the overall accuracy is 88%. At the same time, the classification accuracy of the main land use types such as cropland, rural residence and urban residence are also higher. This study has some practical value to the watershed land use,and it also provides a methodological reference for land use classi- fication of the whole Taihu Lake watershed or other plain river network regions.