基于影像融合的陕北黄土丘陵沟壑区土地利用自动分类

 采用1种遥感影像和单纯的监督分类方法,在黄土丘陵沟壑地区的土地利用调查中,难以获得高精度的土地利用数据。为解决此问题,以陕北无定河流域为研究区,以主成分变换的方法,对多源遥感影像(TM多光谱数据和SPOT全色波段数据)进行融合处理;同时,在分类中,采用监督分类与非监督分类相结合的混合分类法,改进训练样本选取方法,先以非监督分类获得初始训练样本,在对样本进行删除、增补、合并等调整的基础上,再进行监督分类。2种方法的结合使用,使土地利用信息自动提取的精度明显提高。与仅以TM影像为信息源,采用单纯监督分类法的分类结果对比可知:土地利用各类别的提取精度都有不同程度的提高,分类总精度从82.0%提高到89.2%;水体、水田和城镇用地等面积较小类别的精度,提高了10%以上;坡耕地与林草地的混分现象明显减少,精度均提高了5%以上,取得了良好的分类效果。研究结果为陕北黄土丘陵沟壑区土地利用变化动态监测,提供了重要的技术支持和借鉴。

Remote Sensing Classification of Land Use Based on Image Fusion in the Loess Hilly Region of Northern Shaanxi Province

The land use classification accuracy is unsatisfactory based on single remotely sensed data and supervised classification in the land use investigation of loess hill and gully area. Taking the watershed of WuDing River of Northern Shaanxi Province as a test area, the TM multi-spectral data and SPOT pan data are merged by the method of Principal Components Analysis. Then, based on the merged image, the land use categories are extracted by applying an integration of supervised classification and unsupervised classification, which improved sampling method remarkably. The total accuracy increased from 82.0% to 89.2%, especially the accuracy of city and town area, paddy field, water area increased over 10%, the mixture of sloping field and forest (grassland) decreased remarkably and the accuracy of the two categories increased over 5% respectively by the combination of two methods, compared to the classification based on the single TM multi-spectral data and supervised classification. The result is of critical significance in land use dynamic monitoring in the area.