【摘要】 【目的】利用遥感制图对海南省屯昌县土地利用情况进行分析评价。【方法】文章采用海南地区GF-2多光谱4 m和全色1 m影像,经过正射校正、辐射定标、大气校正和融合处理,在eCongnition软件平台上,利用面向对象分类方法,通过构建NDVI、NDWI、SAVI 3种植被指数特征值,结合多尺度分割算法,建立规则集,逐层逐步提取了该地区林地、水体、不透水面和其他植被4种类型的地物。【结果】林地和水体提取结果较优,NDVI对林地和其他植被2个地类识别的贡献率较大。NDWI指数的引入能够较好地解决浅水识别困难的问题,取得了很好的水体分类效果;通过SAVI临界阈值的设定,不透水面和道路能够较好地分类,但是部分裸土地与不透水面容易混淆。分类结果为不透水面12.392 km~2、水体2.534 km~2、林地8.519 km~2、其他植被7.690 km~2。【结论】在eCongnition软件平台的多尺度分割算法和多种指数结合逐层提取土地利用覆盖分类方法具有高效且便于操作等特点,适合在区域尺度上推广。更多还原

【Abstract】 【Objective】To analyze and evaluate the land use in Tunchang County of Hainan Province by using remote sensing mapping.【Methods】The GF-2 multi-spectral 4-meter and panchromatic 1-meter images in Hainan were used to extract the land cover types.After the data preprocessing,the vegetation index NDVI,NDWI and SAVI wer constructed on eCongnition software platform as well as the rule set by using object-based classification approach,combining with multi-scale segmentation algorithm.And then four types of land objects,i.e.woodland,water body,impervious surface and other vegetation were gradually extracted from the image.【Results】The results of extracting forest land and water body were better,and NDVI contributes more for the recognition of forest land and other vegetation.The introduction of NDWI index could solve the difficult problem of shallow water identification in urban areas and surrounding areas,and had achieved good classification results.Through setting SAVI threshold,urban impervious surface and road could be classified better,but some bare land and impervious surface were easy to be confused.The classification results were 12.392 km2 of impervious surface,2.534 km2 of water,8.519 km2 of woodland and 7.690 km2 of other vegetation respectively 【Conclusion】The multi-scale segmentation algorithm and multi-index extraction method based on eCongnition software had the characteristics of high efficiency and easy operation,and were suitable for applicating on regional scale.更多还原