摘要
为了准确快速地获取高分辨率影像中橡胶林的分布信息,设计了一种基于纹理特征和多光谱特征的信息提取方法。方法选取合适的植被指数,将多光谱和植被指数的影像进行地统计半方差分析,获得最佳纹理提取窗口并实现各种纹理信息的提取,将纹理信息和光谱信息一起作为参考特征构建地物的分类规则并用C5决策树分类算法实现。选取某高分辨率遥感影像区域对该方法进行验证,橡胶树林提取的生产者精度为81.00%,提取用户精度为82.65%,总精度为83.50%,Kappa系数为0.78。与其他方法分类结果对比表明,本文方法是一种有效的橡胶林提取方法。
In order to accurately and quickly extract the information of rubber woods, a new informa- tion extraction method of rubber woods distribution was designed based on textural features and multi- spectrum features of remote sensing images with high resolution, and the detail process as follow. firstly, choose suitable vegetable index; secondly, acquire the best texture extraction window through semi-variance statistical analysis of the images of vegetable index and multi-spectrum and extract tex- ture information of remote sensing images; at last, build new classification rules based on texture in- formation and spectrum information of remote sensing image with high resolution and realize the new method by using the arithmetic of C5.0 decision tree. The new method was putted in practiced in re- mote sensing images with high resolution of GuangBa farm DongFang city, HaiNan Province. The re- sults showed that the producer's accuracy, user's accuracy and total accuracy of rubber woods is are 81.00%, 82.65%, and 83.50% respectively, and the kappa coefficient is 0.78. The results that com- paring with other classification methods indicated the method is valid for rubber woods identification.
出处
《武汉大学学报(信息科学版)》
EI
CSCD
北大核心
2014年第4期411-416,421,共7页
Geomatics and Information Science of Wuhan University
基金
海南省自然科学基金资助项目(807019)
海南大学2010青年基金资助项目(gnjj1024)~~
关键词
高分辨率影像
橡胶林
信息提取
决策树分类
remote sensing images with high resolution
rubber woods
information extraction
classi-fication by decision tree