img

官方微信

高级检索

中国沙漠 ›› 2015, Vol. 35 ›› Issue (4): 850-856.DOI: 10.7522/j.issn.1000-694X.2015.00082

• 沙漠与沙漠化 • 上一篇    下一篇

基于高光谱影像数据的戈壁表面砾石粒径定量反演潜力评估

曹晓阳, 穆悦, 曹晓明, 冯益明   

  1. 中国林业科学研究院 荒漠化研究所, 北京 100091
  • 收稿日期:2015-01-27 修回日期:2015-03-11 出版日期:2015-07-20 发布日期:2015-07-20
  • 通讯作者: 冯益明(Email: fengym@caf.ac.cn)
  • 作者简介:曹晓阳(1988-),男,山东潍坊人,博士研究生,主要从事遥感技术应用研究。Email: shadowcxy@163.com
  • 基金资助:

    国家公益性行业(林业)科研专项项目(201404304);国家自然科学基金项目(31370708)

Identification of Gravel Size on the Gobi Surface using EO-1 Hyperspectral Data

Cao Xiaoyang, Mu Yue, Cao Xiaoming, Feng Yiming   

  1. Institute of Desertification Studies, Chinese Academy of Forestry, Beijing 100091, China
  • Received:2015-01-27 Revised:2015-03-11 Online:2015-07-20 Published:2015-07-20

摘要:

在新疆哈密境内的噶顺戈壁选取样地,采集不同粒径砾石的光谱数据,分析光谱吸收特征,并利用光谱混合分析技术,以地物光谱为端元从EO-1 Hyperion 高光谱影像中提取了不同粒径的丰度图,分析戈壁表面砾石与高光谱影像的相关关系。结果表明:获取的地物光谱特征显示出粒径的差异对光谱具有明显的影响。所有光谱都展现出在2 250 nm处具有Al–OH 的吸收特性,而且粒径d=41 cm的吸收特性比其他粒径都更显著。而与更小粒径相比,粒径d=53 cm和d=83 cm在480 nm和920 nm处具有较弱的Fe3+ 吸收特性。粒径d=0.8 cm(R2=0.637)、d=3.4 cm(R2=0.687)、d=16.3 cm(R2 =0.644)及d=41 cm(R2=0.622) 与相应的丰度影像具有显著相关性,而粒径d=53 cm(R2=0.181)和d=83 cm(R2=0.167)与相应的丰度影像相关性不显著。EO-1高光谱影像适用于对戈壁区砾石分布特征的确定,在进一步的研究中将高分辨率影像与高光谱影像相结合,可以提高判别精度。

关键词: 高光谱, 光谱混合分析, 地物光谱, 戈壁砾石

Abstract:

The ground spectra of different size levels of gravel of Gaxun Gobi were obtained in Hami, Xinjiang, China. The spectral absorption features were analyzed, and the abundance images were discriminated from the Hyperion imagery using spectral mixture analysis to analyze the correlations between the gravel size levels and the hyperspectral images. The spectral features demonstrated that differences in gravel size levels can have a considerable influence on the obtained ground spectra. All of the spectra exhibited an Al-OH absorption feature at 2 250 nm and this feature was much more pronounced for gravel with diameter d=41 cm than for other gravel size levels. In contrast with the smaller gravels, gravels with d=53 cm and d=83 cm exhibited much weaker Fe3+ absorption at 480 nm and 920 nm. The gravels with d=0.8 cm (R2=0.637), d=3.4 cm (R2=0.687), d=16.3 cm (R2 =0.644), and d=41 cm (R2=0.622) exhibited significant relationships with the corresponding abundance images, whereas very poor correlations were found for gravels with d=53 cm (R2=0.181) and d=83 cm (R2=0.167).These spectral features and correlation results confirm that EO-1 Hyperion images are suitable for determining the distribution of gravels in the gobi region. The high resolution image should be combined with the hyperspectral images to improve the discriminating accuracy in further study.

Key words: hyperspectral, spectral mixture analysis, ground spectra, gobi gravels

中图分类号: