20130549修改--黄.pdfVIP

增强的高分辨率影像形态学建筑物指数的研究 AnEnhancedMorphologicalBuildingIndexforBuildingExtractionfrom High-ResolutionImages Abstract：High-resolution images are important basic data forurban surface features coverage analysis. This study proposed an enhanced morphological building index (EMBI) for automatic building extraction from high-resolution remotely sensed imagery. Firstly we extracted the urban impervious feature, and then EMBI was built based on a multi-scale white top-hat morphology reconstruction operation on the feature, which taking advantage of the relationshipbetweenthephysicalproperties ofbuildings andmorphologicaloperators.Subsequently,theEMBIfeature image combined with the shape characteristics (length-width ratio, area, etc.) completed the final building extraction using a decision tree method. In order to verify the proposed method, the Washington DC Commercial Street high-resolution hyperspectral HYDICE image and Wuhan University two QuickBird images were used. In these experiments, the EMBI algorithm achieved satisfactory results and outperformed the MBI algorithm in terms of accuracies, i.e. the overall accuracy respectively increased by 7.31%, 6.48%, 7.83%, which proved that the EMBI algorithmperformedmorereliability. Key words：high-resolution images; enhanced morphological building index; urban impervious feature; the shape characteristics;decisiontree 摘 要：高分辨影像是城市地物覆盖分析的重要数据基础，本文提出了一种增强的形态学建筑物指数 （EMBI）， 利用该指数和地物的几何形状约束来完成高分辨率建筑物的自动提取。本方法首先提取城市的不透水层特征， 然后通过建立建筑物属性与形态学运算之间的关系得到了EMBI特征图像以增强对建筑物的描述，随后结合 状特征（长宽比，面积等）采用决策树分析的方法完成对建筑物的最终提取。为了验证本文提出的方法，利用 华盛顿特区商业街的航空高分辨高光谱HYDICE影像和武汉大学的两幅QuickBird影像进行实验，实验的精度 对比反映本文算法比MBI算法能获得更好的建筑物提取结果，其总体精度分别提高了7.31％，6.48％，7.83％， 从而表明EMBI算法更可靠。 关键词：高分辨率影像；增强形态学建筑物指数；城市不透水层；形状特征；决策树 中图分类号：P237 1 引言 高空间分辨率遥感影像能够展现大量的地物特征，包括地物的颜色、尺寸、形状、纹理以及物体间的布局 关系，空间几何信息异常丰富，这使得利用高分辨图像对城市建筑物进行精确识别成为可能。然而，高分辨率 图像中同类地物的光谱统计特征存在很大的光谱异质性，这使得仅利用光谱信息提取地物不能得到满意的结果。