摘要
用计算机对遥感影像进行地物类型识别是遥感数字图像处理的一个重要内容,传统的地物分类一般采用MSS、TM和Spot等遥感影像作为数据源。与MSS、TM和Spot等传统遥感影像相比,QuickBird等高分辨率影像数据量大,混合像元减少、地物信息增大,能够被应用于土地分类。在监督分类中,对于达不到精度要求的模板,通常采用重新选择训练区的方法来进行修正,而本文采用目视修改的方法来对监督分类进行补充。本文方法可以改正初次分类中的误分、混分地物,使其归到正确的地物分类中,显著提高了土地分类的精度。为了验证算法的有效性,利用ERDAS IMAGING遥感图像处理软件进行实验和精度评价。实验结果表明,监督分类和目视修改相结合的地物分类方法可以显著提高图像的分类精度。
Using computer to recognize the type of the ground object on the remote sensing image is an important part of the remote sensing digital image processing. The traditional classification of ground object generally uses remote sensing images, such as MSS, TM and Spot, as data source. Compared with MSS, TM, Spot and other traditional remote sensing images, QuickBird and other high-resolution images have advantages of larger amount of data, reduced mixed-pixel, and increased ground object information, and therefore can be used in the land classification. In the supervised classification, the method of reselecting the training zone is usually adopted to revise the template of which the precision is not high enough. But in this paper, we use the visual revision method to supplement the supervised classification. This method can correct the error and the mixed classification of the ground object in the initial classification brining the ground object back to the correct classification, and significantly improve the accuracy of the land classification. To verify the effectiveness of the algorithm, the experiment and accuracy assessment are carried out using ERDAS IMAGING remote sensing image processing software. The experimental results show that the image classification accuracy can be greatly improved by using the ground object classification method which combines the supervised classification and visual revision.
出处
《国土资源信息化》
2009年第5期37-40,48,共5页
Land and Resources Informatization
关键词
监督分类
目视修改
遥感影像
分类模板
精度评价
Supervised classification
Visual revision
Remote sensing image
Classification template
Accuracy assessment
