【摘要】 以2010年HJ-1号卫星CCD数据为主要数据源,采用面向对象分类方法,提取关中平原人工表面信息。在eCognition软件平台的支持下,综合影像光谱信息、纹理、形状和地理空间关系等特征参数,并添加DEM、Slope、TM数据、样本点数据以及特征指数等辅助信息,依据决策树分类思想,进行多尺度分割逐级提取地物类型。结果表明:利用HJ-1号卫星数据提取区域人工表面信息是可行的,采用面向对象的计算机自动分类方法提取人工表面信息的效率明显提高,且精度达到85.90%。更多还原

【Abstract】 Real-time extraction of the artificial surface is important for the management of land resources and the assessment of ecological environment. The object-oriented classification method based on HJ-1 satellite CCD images was used to extract artificial surface information,and Guanzhong Plain,an important commodity grain base located in the middle of Shaanxi Province with intensive human activities,was chosen to test the adaptability of the extraction method. With the support of eCognition software,the image spectrum information,texture,shape index and geographical spatial relation together with other auxiliary information such as DEM,Slope,TM data,sample point data and some characteristic indices were synthesized,and then artificial surface information were extract hierarchically on the basis of the classification algorithm of decision tree. The classification accuracy was 85. 90% and the results indicated that this method was feasible to extract artificial surface information using object-oriented method based on HJ-1 satellite CCD data. Compared with the traditional visual interpretation method,the efficiency of computer automatic classification based on object-oriented technology has been significantly boosted.更多还原