三峡地区主要地类的自动提取方法研究

针对三峡地区地貌类型复杂、土地利用多元化及遥感影像变异大的特征，在完成图像预处理和辅助资料收集的前提下，提出了一种基于样本和辅助资料可交互作业的计算机全自动土地利用类型提取方法。应用专业的图像处理软件eCognition为平台，通过选取样本、寻找最佳波段并生成判别函数来达到提取的目的，在很大程度上减弱了遥感判读中“同物异谱”、“异物同谱”、“边界过渡”等现象对分类结果的影响。分类结果输出为常用的栅格以及矢量格式，方便地实现了与其它栅格和矢量处理软件的兼容。运用自动提取的方法完成了三峡地区主要土地利用类型的自动提取工作并取得了较高的验证精度；在大面积的土地利用调查中应用遥感图像自动分类方法来获取土地利用信息的技术路线和方法方面作了一定探索。

EXPOLORATIONS ON METHOD OF AUTO-CLASSIFICATION FOR THE MAIN TYPES OF GROUND OBJECTS IN THE THREE-GORGE REGION

On the basis of image preprocessing and collection of auxiliary materials, the paper put forward a kind of automatic category method for the land utilization patterns, in order to undertand the characteris- tics of complicated topography,diversification of land utilization,and great variance of remote sensing ima- ges in the Three-gorge Region. Based on the professional image-analysis software-eCognition, the goal was achieved with the classification of selected samples, the selection of the most suitable wave bands, the es- tablishment of identified functions. The decrease in the influence by the same thing with different spec- trums, different things with the same spectrum,and border transitions was also obtained. The results of classification are common raster or vector forms, which can be compatible with other raster or vector anal- ysis software conveniently. The research accomplished the classification of the main patterns of land utili- zation in the Three-gorge Region with higher precision of verification, verifying the exploration of techno- logical route and method to obtain information on land utilization in lager-scale survey by application of au- tomatic classification of the remote sensing images.