引入分维值进行南京市区土地利用分类

The Computer Classification of Land Use by Fractal Dimension Value

采用盒维数法计算研究区各土地利用类型的分维值,得到分维特征图像,并将它和另外几个波段一起参与研究区的监督分类,和不考虑分维值的监督分类比较,在本研究区,精度提高比较显著的是居民地和林地,保持了交通用地和水域的高分类精度,对园地和耕地没有明显的改善效果.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 fractal dimension value of each type of land use is calculated by box - counting technique and the fractal feature image is also derived. Then the fractal feature image and the other three band images are used to proceed supervised classification. Compared with supervised classification without fractal information and the accuracy of residential area and forestry is promoted effectively and the high accuracy of communication and water area is retained. But there isn＇t distinct improvement for garden plot and cultivated area. The result indicates that the fractal information can promote the accuracy of computer classification in some degree.