Landsat7 ETM+图像用作SPOT5图像森林分类的辅助数据研究

为改善面向对象的SPOT5图像分类精度,将Landsat7 ETM+数据作为辅助数据,试验了3种图像分割和对象特征提取方案,在图像分割、基于规则的分类、基于分类的分割和对象特征提取与筛选的基础上,进行了Bayes准则分类试验。结果表明:ETM+图像信息参与分类明显地提高了SPOT5图像的分类精度,但由于空间分辨率低,在森林分布破碎、类型和种类多样、结构复杂的研究区域,ETM+数据只能用于对象特征提取,不能实质性地参与图像分割,否则会导致图像对象边界不准确、同质性降低、异质性增加,影响图像分类精度。

Using Landsat7 ETM+ images as ancillary data for forest cover classification of SPOT5 images

In order to improve the accuracy of object-oriented classification of SPOT5 images, Landsat7 ETM+ images were used as ancillary data in three protocols of image segmentation and object feature extraction. Using Bayes rule, the classification of images was performed, followed by image segmentation, rule-based classification, classification-based segmentation, object-feature extraction and selection. The results indicate that the application of ETM+ data can improve the classification accuracy of SPOT5 images. However, with low spatial resolution, this protocol can only be used in the extraction of object features, and not in image segmentation in regions with fragmented forests, diverse species, types, and complex structures. Such conditions lower the precision of object borders, reduce homogeneity and increase the heterogeneity of objects, which in turn affect the classification accuracy.