【摘要】 新增建设用地及其占用地类是年度土地利用动态遥感监测的重点关注信息,综合利用多源数据是提高占地信息提取精度的主要手段之一。现有研究基于光学与高光谱特征波段的像素级融合影像开展分析,不可避免地导致了高光谱影像光谱信息的损失。为了尽量避免信息损失,提出一种光学和高光谱影像独立分析、结果集成的新增建设用地占用地类信息提取方法。该方法首先结合土地利用动态遥感监测所定义的3种主要新增建设用地形态,利用前后两个时相GF-1光学影像的光谱和纹理特征提取新增建设用地,其次利用前时相Hyperion高光谱影像的精细光谱成像能力进行详细土地利用分类,最后通过新增建设用地与土地利用分类图的叠置分析提取新增建设用地占用地类。在北京市丰台区典型区域开展的验证实验取得了86.71%的新增建设用地占用地类属性精度,表明本方法可为土地利用动态遥感监测提供较为准确的地类变化信息。更多还原

【Abstract】 Land occupied by newly increased construction land is an important information concerned by land-use dynamic monitoring using remote sensing. As well known that combined utilization of multi-source data is one of the main methods to improve accuracy of land-use monitoring, however, most of state-of-the-art researches carry out analysis on the pixel-level fused image of visible image and selected bands drawn from hyperspectral image. To avoid the spectral information loss of hyperspectral image this paper proposes a three-fold method which analyzes visible and hyperspectral independently and then combines the analysis results to extract land occupied by newly increased construction land.Firstly, three main types of newly increased construction land in land-use monitoring are introduced, which is then extracted based on the spectral and textural features drawn from bi-temporal visible images. Secondly, detailed land-use mapping is conducted on hyperspectral image of former temporality. Finally, the land-use mapping result is overlaid with the newly increased construction land result to extract occupied land information. Experiments carried out in Fengtai district, Beijing indicate that attribute accuracy of occupied land is up to 86.71%, which proves that the proposed method is able to provide land-use changing information for land-use dynamic monitoring with reliable accuracy.更多还原