基于分层次多尺度分割的面向对象地物分类方法

Object-oriented Classification Method for Surface Features Based on Hierarchical and Multi-scale Segmentation

高分遥感影像具有数据量极大、数据复杂以及尺度依赖的特点,如何从影像中提取更精准、更高质量的地物信息一直都是国内外科研人员的关注热点。通过研究基于分层次、多尺度分割的面向对象技术,并运用该技术对高分影像中的地物信息进行分类,通过试验对分类结果做了基于误差矩阵的精度评价,其分类结果的Kappa系数达到0.918,验证了该方法的可行性,为以后自动选取最佳分割尺度与参数的研究提供了参考。

High-resolution remote sensing image is of huge data volume, complicated and scale-dependent. Extracting more accurate and higher quality ground object information from the image has always been the focus of researchers at home and abroad. In this paper, the object-oriented technology based on multi-level and multi-scale segmentation is studied, and the technology is used to classify the ground information in high-resolution images. Finally, the accuracy of the classification results based on the error matrix is evaluated through experiments, and the Pascal coefficient reaches 0.918, which verifies the feasibility of the method. And this will provide reference for the study of selecting automatically the optimal segmentation scale and parameters.