首页 | 本学科首页   官方微博 | 高级检索  
     

基于eCogniton的高分辨率遥感图像的自动识别分类技术
引用本文:王文宇,李博.基于eCogniton的高分辨率遥感图像的自动识别分类技术[J].北京建筑工程学院学报,2006,22(4):26-29.
作者姓名:王文宇  李博
作者单位:1. 北京建筑工程学院,测绘与城市空间信息学院,北京,100044
2. 北京师范大学,地理学与遥感科学学院,北京,100875
基金项目:北京建筑工程学院博士启动基金(1005003)致谢:本研究由北京建筑工程学院博士启动基金资助,感谢航天星图科技(北京)有限公司李静提供的遥感例子数据,以及给予的技术指导.
摘    要:传统的遥感信息分类和提取,主要是利用数理统计与人工解译相结合的方法.这种方法不仅精度相对较低,效率不高,而且依赖参与解译分析的人,在很大程度上不具备重复性.专业高分辨率遥感影像分类软件eCogniton采用一种全新的面向对象图像的分类技术来进行影像的分类和信息提取.面向对象图像分类技术的关键技术在于:(1)用来解译图像的信息并不在单个像元中,而是在图像对象和其相互关系中;(2)eCogniton采用多分辨率对象分割方法生成图像对象,提高了分类信息的信噪比;(3)基于对象的分类技术不同于纯粹的光谱信息分类,图像对象还包含了许多的可用于分类的一些其他特征:形状、纹理、相互关系、上下关系等信息.eCogniton的分类结果与传统分类方法相比。其特征提取算子更加地适合于几何信息和结构信息丰富的高分辨率图像的自动识别分类.

关 键 词:eCognition  高分辨率遥感图像  自动识别分类  面向对象
文章编号:1004-6011(2006)04-0026-04
收稿时间:2006-10-27
修稿时间:2006年10月27日

Auto-identify Classification Technology Adopted to High-resolution Remote Sensing Image by eCognition Software
Wang Wenyu,Li Bo.Auto-identify Classification Technology Adopted to High-resolution Remote Sensing Image by eCognition Software[J].Journal of Beijing Institute of Civil Engineering and Architecture,2006,22(4):26-29.
Authors:Wang Wenyu  Li Bo
Affiliation:1. School of Geomatics and Urban Information, Beijing 100044 2. School of Geography and Remote Sensing, Beijing Normal University, Beijing 100875
Abstract:The traditional classification and extraction of the information about remote sensing mainly make use of the method combining science statistic with man-made interpretation,which is low in precision and inefficient.Meanwhile,the method leaning on the people who participate in the interpretation and analysis doesn't have repetition to some extent.Special high-resolution remote sensing classification software eCognition adopts a classification technology of the bran-new object-oriented to classify the image and extract information.The key technology of object-oriented image classification is that: the first,the information being used to interpret the image doesn't exist in single pixel but in the interrelation of image and the pixel.The second,eCognition adopts the method of multi-resolution object cutting to produce the image object,and it improves the SNR of the classification information.The third,the classification technology based on object is different from the pure spectral information classification,the image object also includes many other features used to classify information such as shape,vein and interrelation and so on.A comparison of the classification results of eCognition and the traditional classification method indicates that the feature extraction operator of eCognition is more adapt for auto-identify classification of high-resolution remote sensing whose information is abundant in structure and geometry than the traditional classification method.
Keywords:eCognition  high-resolution remote sensing  auto-identify classification  object-oriented
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号