机构地区: 武汉大学遥感信息工程学院
出 处: 《地理空间信息》 2015年第4期86-88,101,共4页
摘 要: 从高光谱影像能够提供地物连续光谱曲线的优势出发,提出了提取地物诊断性光谱吸收峰的特征参数进行地物精细分类的方法。用OMIS高光谱数据进行实验,首先对光谱曲线进行包络线去除处理,然后在归一化的曲线上提取光谱吸收峰的形态特征参数,根据不同种地物的光谱差异与分类需求进行特征参数选择,用于地物分层精细分类,在每一类别的地物之中实现不同子类的区分。分类总体精度达到81.022 6%,Kappa系数为0.748 9,尤其在植被和水体的子类区分上取得了较好的效果,证实了该方法的有效性。 A novel method based on the extracting of the diagnostic spectral absorption peaks of object was proposed to make a refined classification of hyperspectral image in this paper.OMIS data was used in this experiment.The morphological characteristic parameters of continuum removed spectrum curve were extracted to classify ground objects.Different characteristic parameters were chosen in the hierarchical classification,and different subclasses were divided in each class.The overall classification accuracy was 81.022 6%,kappa coefficient was 0.748 9,and the subclasses of plants and water bodies were divided well.The result indicated the effectiveness of this method.