摘要
以长白山金仓林场的TM卫星影像为数据源,将采用常规的计盒维数法计算得到的分维特征图像参与研究区的监督分类,和无分维信息的监督分类相比较,阔叶林的分类精度提高显著,非林地中的农田和建筑用地分类精度有一定提高,针阔混交林和针叶林的分类精度提高不明显。结果表明,加入分维信息可以在一定程度上提高图像监督分类的精度。
In this paper the TM images of Jincang forestry center in Changbai mountain were used, and the fractal feature image calculated by box-counting technique was used to proceed supervised classification. Compared with supervised classification without fractal information, the accuracy of broadleaf tree was promoted effectively; the accuracy of residential area and farmland was promoted in some degree, but there wasn＇t distinct improvement for conifer and mixed forest. The result indicated that the fractal information can promote the accuracy of forestry classification in some degree.
出处
《安徽农业科学》
CAS
北大核心
2005年第10期1889-1890,共2页
Journal of Anhui Agricultural Sciences
关键词
计盒维数法
监督分类
分维值
Box-counting technique
Supervised classification
Fractal dimension value
