| 摘要: |
| 以黄土高原天然林为研究对象，以TM（专题制图仪）影像为主要数据源，结合森林资源二类调查数据，对该地区天然林林地类型信息的提取方法进行了研究。首先对TM影像的6个波段（除热红外波段6）进行主成分变换，以减小特征间的相关性，并将变换后影像的第一、第二和第三主成分分量进行RGB（红、绿、蓝通道）合成，这样得到的影像可以更好地表达地面信息；然后参考先验知识纯化样本，以建立比较精确的分类器，在此基础上进行计算机监督分类，获取了研究区域各地类的面积。结果表明，最终分类精度可以达到92％，Kappa值达到0．906。说明在该地区天然林区林地信息的提取中，利用TM作为数据源，经过图像处理和监督分类能够达到较高的精度，可以满足林业生产的要求。 |
| 关键词: 遥感 信息提取 计算机分类 |
| DOI： |
| 分类号: |
| 基金项目:国家林业局天然林保护工程科技支撑项目(TBKJ2003-05) |
|
| Study on extracting forestland information of natural forest based on RS and computer classification in Loess Plateau |
|
|
| Abstract: |
| In this research,the TM image is used in the PCA transformation in order to reduce the correlation among the features.The new image is synthesized by the first,second,third principal components in the way of RGB after PCA.It is obvious that new image is better in expressing the ground information.With the reference knowledge the samples were purified and a more precise classification signature was built.Then,the classification was supervised and areas of each type were computed.Through evaluating classification accuracy,the results show that the method above is proper and the precise achievement can be obtained by using it.The final data can be used to update the information,and it can provide precise data for making the thematic map of the forest. |
| Key words: RS TM information extract computer classification |
