摘要
高分辨率遥感影像分类一直是业内研究的热点之一,考虑到影像地物光谱角和光谱距离在分类中具有较好的互补性,提出了一种基于光谱角和光谱距离自动加权融合的分类方法,对传统多分类器分类的融合策略进行改进,能够在训练阶段根据样本自动地调整好各分类器对各类别进行分类的权重系数,使得融合后的分类结果更加科学和准确。QuickBird影像的分类实验表明,方法的分类精度明显优于单纯的光谱角或距离法,可广泛用于各种高分辨率影像的分类识别。
High resolution remote sensing image classification was one of the vital research points in the remote sensing ficht. Considering the high complementary of spectral angle and spectral distance in the classification, a new method called autonmtie weighting fusion classification based on spectral angel and spectral distance was proposed. It was an improvement of the strategy to merge the results of different classifiers based on automatic weighting fusion for different classifiers, which promoted the classification accuracy. The experimental results of QuickBird images showed that the classification accuracy of this method obviously exceeded both spectral angel classification and spectral distance classification, and this meLhod could be widely used to classify and recognize various high spatial resolution remote sensing images.
出处
《地质学刊》
CAS
2012年第1期33-36,共4页
Journal of Geology
基金
国家高科技研究发展计划(2007AA12Z156)
国家自然科学基金(40672195
41072245)
北京市自然科学基金(4102029)
