结合目标分割的高光谱城市地物分类

发布时间：2018-12-16 17:33

【摘要】：结合高光谱影像地物光谱特征与高空间分辨率影像分割获得的目标对象进行地物分类。首先,对Hyperion影像进行坏线和Smile效应去除,经过FLAASH大气校正后,得到研究所用的155个波段;其次,利用地物光谱曲线的特征点确定适合地物识别的光谱分辨率,进行Hyperion影像降维,生成降维后所需的21个宽波段;然后,对IKONOS影像采用小波融合,利用多分辨率分割技术生成高空间分辨率影像目标对象;最后,基于层次分析法对分割后生成的目标对象进行分类,采用模糊隶属函数利用植被红边效应、水体在近红外波段吸收特征进行第1层次分类,再取距离值最大的前10个Hyperion影像波段作为标准最邻近分类的特征波段,完成第2层次分类。分类结果表明,研究区共分出9种地物类型,分类效果明显优于最大似然法分类与光谱角填图法。
[Abstract]:Combined with the spectral features of hyperspectral images and the segmentation of high spatial resolution images, the objects are classified. Firstly, the bad line and Smile effect are removed from Hyperion image, and 155 bands are obtained after FLAASH atmospheric correction. Secondly, the spectral resolution suitable for ground object recognition is determined by using the characteristic points of the spectral curve of the ground object, and the dimension reduction of Hyperion image is carried out, and the 21 wide bands required after dimension reduction are generated. Then, the IKONOS image is fused by wavelet, and the object of high spatial resolution image is generated by multi-resolution segmentation technology. Finally, based on the analytic hierarchy process (AHP), the target objects generated after segmentation are classified by fuzzy membership function using the red edge effect of vegetation, and the water body is classified at the first level in the near infrared band absorption characteristics. The first 10 bands of Hyperion images with the largest range are taken as the feature bands of the standard nearest neighbor classification, and the second level classification is completed. The results show that there are 9 types of ground objects in the study area, and the classification effect is obviously superior to that of maximum likelihood method and spectral angle mapping method.
【作者单位】：闽江学院地理科学系;
【基金】：国家自然科学基金项目“基于MODIS BRDF产品的叶片聚集度系数遥感反演与验证”(编号:41271354) 福建省科技厅资助项目“基于高光谱特征与目标分割的城市地物识别研究”(编号:2015J01627) 闽江学院资助项目“摄影测量学实践教学改革”(编号:MJU2014BD19)共同资助
【分类号】：TP79

【相似文献】