基于面向对象的香榧资源分布遥感调查研究  

Survey of Torreya Grandis Merrillii Resource Based on Object-oriented Method

作　　者：王新辉[1] 沈掌泉[1] 王珂[1] 贾春燕[1] 凌在盈[1] 

机构地区：[1]浙江大学农业遥感与信息技术应用研究所,杭州310029

出　　处：《科技通报》2009年第2期160-166,共7页Bulletin of Science and Technology

基　　金：国家自然科学基金资助(30671212)

摘　　要：香榧系第三纪孑遗植物,为我国特有的珍贵经济树种。常规的香榧资源调查方法存在工作量大、数据时效性差等不足。近年来高分辨率遥感影像的应用,为特定树种信息提取提供了可能。基于此,本文以高分辨率IKONOS卫星影像为基础,采用面向对象的信息提取方法,多尺度分割形成对象后,利用光谱、形状、纹理等构建特征空间,进行会稽山区香榧信息提取的试验,并与常规监督分类法(最大似然法)进行了比较。结果显示,基于面向对象方法的香榧信息提取精度达到86.57%,比监督分类法的精度提高了27.90%。研究表明,用面向对象的分类方法进行香榧信息提取和资源调查是可行的。Torreya grandis Merrillii is tertiary relict plant, and is a precious economic species peculiar in China. The conventional plant resource survey is time-consuming and cost-expensive,so it is difficult to update in time. The high spatial resolution images have more spatial details and texture characteristics,and offer the possibility to extract information of a particular tree species. In this study,an object-oriented technology was introduced to extract the information of Torreya grandis Merrillii from IKONOS image. Firstly, the image was segmented at three scales and objects were formed; secondly,the feature space was built using the characteristics of spectrum, shape and texture; and finally ,a combining of near neighboring and fuzzy expert system was used to identify Torreya grandis Merrillii trees. The result showed that the classification accuracy for Torreya grandis Merrillii reached 86.57% ,higher than maximum likelihood classification by 27.90%.

关 键 词：遥感 高分辨率影像 面向对象分类 香榧 多尺度分割 

分 类 号：P96[天文地球—自然地理学]