摘要
利用面向对象的信息提取技术,以高分辨率的广州市QuckBird影像为例,将城市用地分为:居民地、水体、道路、林地和农业用地等5类,并将其与传统基于像素光谱信息的分类方法进行了比较。结果表明:视觉上,面向对象的分类方法克服了传统方法无法克服的"椒盐"噪声的影响;精度上,面向对象信息提取技术的总体精度高达89.53%,比传统方法提高了11%;并且各类地物信息的提取精度均有所提高,其中林地、道路的精度有了较大提高。
： In this paper we compare the performance of two image classification paradigms （object- and pixel-based） for creating a land cover map of suburb of Guangzhou city, China using High-resolution satellite image, the QuickBird images. The image was classified into residential area, water, road, and forest and agriculture land by using supervised and unsupervised combined classification for the pixel-based approach and nearest neighbor （NN） method for the object-oriented approach. The classification outputs were assessed using overall accuracy and Kappa indices. The pixel- and object-based classification methods result in an overall accuracy of 78.53% and 89.53%, respectively. The Kappa coefficient for pixel- and object-based approaches was 0.73 and 0.87, respectively. The object-oriented method greatly lighten the noise influence, has higher classification accuracy and efficiency than that achieved by pixel-based method. Meanwhile, the classification result of object-oriented method is much easier to understand and explain.
出处
《地理空间信息》
2009年第3期62-65,共4页
Geospatial Information
基金
国家自然科学基金资助项目(40801182)
教育部科学技术研究基金资助项目(108162)
中国地质大学(武汉)优秀青年教师基金资助项目(CUGQNL0823)
关键词
高分辨率卫星影像
面向对象
基于像素
精度
high-resolution satellite image
object-oriented
pixel-based
accuracy
