基于天绘一号卫星影像的水体信息提取对比研究

针对天绘一号卫星高分辨率影像,采用面向对象分类方法对怀柔水库区域进行水体信息提取,在多尺度分割的基础上,统计地物的光谱信息、形状因子和亮度均值等,建立水体信息的特征集,充分利用高分辨率的特点提取水体信息,同时选取了参数相近的SPOT和RapidEye两幅国外高分影像进行对比研究,使用相同方法进行水体提取,对实验过程和结果进行了对比分析。针对提取结果,采用野外采样和矢量图分析两种方法综合进行精度评价,根据采样数据得到的精度分别为96.97%,95.45%,92.42%,分析实验结果的矢量图,其中天绘影像水体提取面积为5 537 412.5㎡,SPOT影像为5 398 225㎡,RapidEye影像为5 053 262.5㎡,对实际水域的面积覆盖分别达到了101.40%,98.85%,92.54%,天绘影像的整体精度较高,但在细节表现上较为模糊,主要误差来自于对湿地的误分。实验制定了适用于天绘影像的水体提取方法和规则,分析不同因素对分割与分类结果的影响,同时,我们比较了天绘影像与国外同级别高分影像的优劣性,为天绘影像的进一步应用提供了参考。

The Comparative Research of Water Body Extraction Based on Mapping Sateilite-1 Imagery

Mapping Satellite-1 is the first transmission-type stereo mapping satellite applied for both military and civilian in China, it is carrying with high resolution panchromatic sensor（resolution 2m） and 4-band multi- spectral sensor（resolution 10m）. Based on Mapping Satellite-1 high resolution imagery, we choose Huairou Reservoir（Beijing） as the experimental zone. Object-oriented information extraction technology was used to extract water bodies, based on the result of multi-resolution segmentation, we count spectrum, appearance factor and brightness of features, after analysis and observation, then set the rules to extract water bodies. For comparative research, we also select 2 same level foreign satellite（SPOT and RapidEye） images as comparative data, use the same method to extract water bodies, analysis the difference in both experiment process and results among three images. Finally, we went to the experimental field, analyzed the vector map to access accuracy about the extraction results. Based on the field samples, the water extraction accuracy of Mapping Satellite-1 image, SPOT and RapidEye was 96.97%, 95.45%, 92.42% respectively, compared with the vector map, the water area extracted is 5537412.5m^2（Mapping Satellite-1 Image）, 5398225m^2（SPOT） and 5053262.5m^2（RapidEye）, the coverage of the actual water was up to 101.40%, 98.85%, 92.54% respectively. The accuracy of Mapping Satellite-1 Imagery gets better performance in the water extraction, but its details is a little fuzzy in the segmentation, and its main error is misidentification from wetland to water. Generally speaking, the study formulated water extraction method and rules adapted tfrom Mapping Satellite-1 image, and analysis influence factors for the images＇segmentation and classification. The advantages and disadvantages of Mapping Satellite-1 image compared to the same level foreign satellite images are given.