论文部分内容阅读
及时获取有效的土地覆盖信息是地球系统模拟的基础。因此,中等空间分辨率传感器如MODIS或MERIS空前的通道设置与观测能力,使其具有快速更新土地覆盖图的能力。本文说明了如何结合MERIS的空间维(像元大小为300m)、光谱维(可见光与近红外范围内15个通道)和时间维(重返周期2—3d),用于获取不同区域土地覆被组分的亚像元级组成权重。利用4月、7月和8月三期MERIS FR1b级数据得到荷兰主要土地覆被类型的组成权重。单一时相和多时相的数据都使用单个像元最优化的端元数进行线性光谱分解。利用一种形态偏离指数得到MERIS的空间维并用于端元的选择。应用荷兰土地利用数据库(LGN5)25m分辨率的栅格数据作为本文的参考数据。基于这种数据的高分辨率,因此可以从像元和亚像元的水平同时评价的分类精度。结果显示,结合4月和7月的影像可以获得最优的分类结果,精度约为58%。总的说来,亚像元和像元级的分类精度相似。通过几种组分类别和日期的光谱融合表明,物候状况对于数据获取时相最佳结合的选择以及正确识别土地覆盖类型的重要性。
Timely access to effective land cover information is the foundation of earth system modeling. As a result, unprecedented spatial setup and observation capabilities of mid-spatial resolution sensors such as MODIS or MERIS give them the ability to rapidly update land cover maps. This article explains how to combine MERIS spatial dimensions (cell size of 300 m), spectral dimensions (15 channels in the visible and near infrared range) and time dimensions (return cycles of 2-3 d) for acquiring land cover Sub-pixel level component weights of the components. Using the April, July and August MERIS FR1b data, the weight of the composition of the main land cover types in the Netherlands is obtained. Single-phase and multi-phase data are linearly decomposed using the single-cell optimized end-points. The MERIS space dimension is derived using a morphological deviation index and used for the selection of endmember. The raster data with a resolution of 25m from the Netherlands Land Use Database (LGN5) is used as a reference for this article. Based on the high resolution of this data, the classification accuracy can be simultaneously evaluated from the pixel and sub-pixel levels. The results show that combining the images of April and July can get the best classification results with the accuracy of about 58%. In general, sub-pixel and pixel-level classification accuracy is similar. The spectral fusion of several component categories and dates shows the importance of the state of phenology for the optimal combination of phases in data acquisition and the correct identification of land cover types.