不同高分辨率影像的多尺度叠加分类研究
2021-07-27石芳华
石芳华



摘 要:随着传感器技术的不断发展,高分辨率遥感影像数据的获取也越来越方便。但在遥感影像分类中,并不是空间分辨越高,分类精度就越高。以SPOT5 HRG影像和高分一号影像为数据源,进行多尺度分类实验。首先,采用立方卷积法对两种影像进行尺度扩展,利用变异函数计算影像中各地类的最优尺度;其次,采用最大似然法对其进行多尺度分类;最后,利用混淆矩阵、总体分类精度和Kappa系数对其分类精度进行评价。结果表明,研究多尺度分类方法可以提高遥感影像的分类精度。
关键词:多尺度 最优尺度 高分辨率影像 遥感分类
中图分类号:P237 文献标识码:A文章编号:1672-3791(2021)03(a)-0087-03
Multi scale overlay classification of different high resolution images
SHI Fanghua
(Hubei University of technology, Wuhan, Hubei Province, 430064 China)
Abstract: Abstract: With the development of sensor technology, the acquisition of high-resolution remote sensing image data is more and more convenient. But in the classification of remote sensing images, the higher the spatial resolution is, the higher the classification accuracy is. The multi-scale classification experiments were carried out with SPOT5 HRG image and High Resolution NO. 1 image as data sources. Firstly, the cubic convolution method is used to expand the scale of two kinds of images, and the optimal scale of different classes in the image is calculated by using the variation function. Secondly, the maximum likelihood method is used to classify them in multi-scale. Finally, the classification accuracy is evaluated by using confusion matrix, overall classification accuracy and kappa coefficient. The results show that the multi-scale classification method can improve the classification accuracy of remote sensing images.
Key words: multi scale; optimal scale; high resolution image; remote sensing classification
更高空間分辨率遥感影像数据的获取,随着传感器技术的不断发展也成为可能。人们常常以为采用空间分辨率高的遥感影像,能提高土地覆盖分类的分类精度。其实不然,即使是高空间分辨率影像也很难反映真实客观的地学特征现象。一般来说,遥感信息都存在尺度效应[1]。因此,在研究特定的目标时,需找到一个合适分辨率的遥感信息,反映研究目标的空间分布结构特性。
目前为止,最佳分辨率(最优尺度)的选择方法都是基于统计学理论。Woodcock和Strahler提出了一种用遥感图像的平均局部方差来确定最优分辨率的方法,当局部方差值达到最大时,对应的空间分辨率被称为最优的空间分辨率。Atkinson等通过计算不同空间分辨率影像的变异函数来确定最优分辨率,变异函数值达到最大时,对应的空间分辨率即为最优空间分辨率[2]。……
