协同WP-GS融合、导数变换、多尺度分割和多谱段SAM的分层波谱识别—以兰州、榆林EO-1/Hyperion高光谱图像为例
投稿时间:2018-07-10  修订日期:2019-01-15  点此下载全文
引用本文:
摘要点击次数: 235
全文下载次数: 0
作者单位E-mail
刘 炜 西藏光信息处理与可视化技术重点实验室 remote2009@126.com 
基金项目:国家自然科学基金项目(面上项目,重点项目,重大项目)——山地城市景观格局空间异质性与热环境的多尺度响应特征分析与数值模拟
中文摘要:【目的】提出分层波谱识别法,分别从兰州、榆林市Hyperion高光谱图像上识别9种目标地类,并与SVM监督分类对比。【方法】针对Hyperion图像波谱识别的4个难点:光谱信息高保真融合、敏感谱段提取、“椒盐效应”去除、 消除“同物异谱” 现象导致的误判,设计WP-GS融合、导数变换、4尺度面向对象分割和多谱段SAM协同应用,解决上述难点;基于Hyperion导数变换图像分析波谱变化特征,提取敏感谱段,从4个尺度层依次识别9钟目标地类。并基于目视评判和定量评价,与采用Gram-Schmidt光谱锐化方法融合/Savitzky-Golay卷积滤波/PCA变换的SVM监督分类,比较识别精度。【结果】WP-GS融合方法的光谱保真效果优于Gram-Schmidt光谱锐化方法;4尺度面向对象分割抑制“椒盐效应”的效果优于Savitzky-Golay卷积滤波、移动均值滤波;多谱段SAM利用导数波谱特征能够消除因照度不同对同一类别地物的误判。采用分层波谱识别法,兰州市Hyperion图像波谱识别的总体精度、Kappa系数分别为89.52%、0.852,较SVM分类分别提高18.68%和17.52%;榆林市Hyperion图像识别地物的总体精度、Kappa系数分别为91.12%、0.873,较SVM分类分别提高17.80%和16.89%。【结论】分层波谱识别法协同应用多种技术一体化解决Hyperion图像应用难点,有效利用导数波谱变化特征提取目标敏感谱段,在复杂环境下识别目标地类的能力优于SVM监督分类。
中文关键词:星载高光谱图像  导数波谱特征  敏感谱段  同物异谱  多尺度  多谱段SAM
 
Hierarchical spectrum recognition based on WP-GS image fusion, derivative transformation, object oriented segmentation and multispectral SAM —Take EO-1/Hyperion images of Lanzhou and Yuyang as an example
Abstract:【Objective】This paper points out hierarchical spectrum recognition in the aspect of hyper-spectral image classification using Hyperion images of Lanzhou and Yuyang in China, by the comparison study of SVM supervised classification.【Method】There are 4 problems for hyper-spectral image classification: spectral information fidelity image fusion, extraction of sensitive band, removal of "salt and pepper effect", avoid misclassification for "same body with different spectrum" phenomenon. Hierarchical spectrum recognition puts forward 4 methods for solving these issues, which are WP-GS image fusion, derivative transformation, object oriented segmentation with 4 scales, multispectral SAM. Hierarchical spectrum recognition can identify 9 kinds land type exactly, based on sensitive bands extracted from derivative transformation image. Visual examination and quantifiable evaluation have been executed to verify authenticity. By contrast, SVM supervised classification was tested, with Gram-Schmidt Spectral sharpening/ Savitzky-Golay convolution filtering/PCA transformation.【Result】Using the method of hierarchical spectrum recognition proposed in this paper, the above 4 problems for hyper-spectral image classification were solved, and good classification results was achieved, with overall classification accuracy and kappa coefficient as 87.13%, 0.8303 in Lanzhou, 87.13%、0.8303 in Yuyang. Growth of 15.97% in Lanzhou and 87.13% in Yuyang for overall classification accuracy, as well as 12.44% in Lanzhou and 87.13% in Yuyang for kappa coefficient were achieved compared with SVM supervised classification.【Conclusion】Compared with SVM supervised classification, hierarchical spectrum recognition proposed in this paper can provide more accurate recognition results for Hyperion images classification.
keywords:satellite-borne hyper-spectral image  derivative spectrum feature  sensitive bands  same body with different spectrum  multi-scale  multispectral SAM
查看全文  查看/发表评论  下载PDF阅读器

版权所有:《红外与毫米波学报》编辑部