密度分割和面向对象结合的东平湖湿地信息提取

Information Extraction from Dongping Lake Wetland with the Density Segmentation and Object-oriented Method

针对湿地复杂的覆被信息,以东平湖湿地为研究对象、Landsat ETM+影像为数据,通过密度分割、设置阈值精确提取植被类型;基于面向对象,结合形状和纹理特征,精确提取水体类型;采用最大似然法提取建设用地及其他地物。结果表明,2009年东平湖湿地面积179.84 km^2,其中自然湿地161.32 km^2,人工湿地18.52 km^2。密度分割和面向对象结合提取的总体精度为88.54%,总体Kappa系数为0.8776,分别高于最大似然法的10.10%和0.1072。湿地信息提取结果可为东平湖湿地管理提供基础数据。密度分割与面向对象结合的湿地遥感信息提取方法可为湖泊湿地快速、准确提取提供技术支撑。

Because wetland information has the characteristics of complexity and diversity in the lake region, it is difficult to acquire accurate information from remote sensing image. Selecting Dongping Lake wetland as the research object, using Landsat ETM+ image data, the paper extracted vegetation types accurately through density slicing and threshold setting;water types were extracted accurately based on the object-oriented method combined with shape and texture characteristics;construction land and other types were extracted with the maximum likelihood method. The results indicated that the wetland area of Dongping Lake was 179.84 km2 in 2009, with the area of natural wetland 161.32 km2, and artificial wetland 18.52 km2. The overall accuracy of density segmentation and object-oriented method was 88.54% and the overall Kappa coefficient was 0.8776, which were higher than 10.10% and 0.1072 compared with the maximum likelihood method. Wetland information extraction results can provide basic data for Dongping Lake wetland management. Wetland remote sensing information extraction method of density segmentation combined with object-oriented method can provide technical support for rapid and accurate extraction of lake wetland.