自然资源学报 ›› 2013, Vol. 28 ›› Issue (7): 1243-1254.doi: 10.11849/zrzyxb.2013.07.016

• 资源研究方法 • 上一篇    下一篇

高光谱植被覆盖度遥感估算研究

包刚1,2,3, 包玉海1, 覃志豪2, 周义2, 黄明祥1,4, 张宏斌3   

  1. 1. 内蒙古师范大学 内蒙古自治区遥感与地理信息系统重点实验室, 呼和浩特 01002;
    2. 南京大学 国际地球系统科学研究所, 南京 21009;
    3. 呼伦贝尔草原生态系统国家野外 科学观测研究站, 北京 100081;
    4. 环境保护部信息中心, 北京 100029
  • 收稿日期:2012-07-19 修回日期:2012-12-06 出版日期:2013-07-20 发布日期:2013-07-20
  • 通讯作者: 包玉海(1965-),男,内蒙古科右中旗人,教授,博士,主要从事土地利用/覆盖变化、自然灾害遥感监测与风险评估研究。E-mail:baoyuhai@imnu.edu.cn E-mail:baoyuhai@imnu.edu.cn
  • 作者简介:包刚(1978-),男,蒙古族,内蒙古库伦旗人,博士研究生,主要从事遥感与地理信息系统应用研究。E-mail:baogang@imnu.edu.cn
  • 基金资助:

    国家重点基础研究发展计划(973计划)项目(2010CB951504);国家自然科学基金项目(41161060);内蒙古自然科学基金项目(2012MS0607);内蒙古师范大学"十百千"人才计划项目。

Hyper-spectral Remote Sensing Estimation for the Vegetation Cover

BAO Gang1,2,3, BAO Yu-hai1, QIN Zhi-hao2, ZHOU Yi2, HUANG Ming-xiang1,4, ZHANG Hong-bin3   

  1. 1. Inner Mongolian Key Laboratory of Remote Sensing and Geographic Information System, Inner Mongolia Normal University, Huhhot 01002;
    2. International Institute for Earth System Science, Nanjing University, Nanjing 21009;
    3. Hulunber Grassland Ecosystem Observation and Research Station, Beijing 100081, China;
    4. Information Center of Ministry of Environmental Protection, Beijing 100029, China
  • Received:2012-07-19 Revised:2012-12-06 Online:2013-07-20 Published:2013-07-20

摘要:

以北京大学"无人机遥感载荷综合试验场"为试验区,采集草地植被覆盖度(Vegetation Cover, VC)和相应样方冠层高光谱反射率数据,并对比研究了高光谱反射率三种变换形式(小波能量系数、主成分和植被指数)与VC之间的关系模型。结果表明:在三种变换形式中,植被指数模型(R2大于0.8,RMSE小于等于0.018 8)优于基于小波变换和主成分分析的VC模型;经过对高光谱数据进行小波分解获得的第二和第四小波能量系数与VC之间存在显著的对数相关(R2分别为0.811和0813;RMSE分别为0.019 9和0.019 8);以多个小波能量系数作为自变量的VC多元回归模型明显优于基于主成分的多元线性回归,R2RMSE分别提高0.058和0.030;将高光谱EVI模型与TM-EVI数据相结合生成的试验区VC空间分布总体上呈北部和南部植被覆盖度高(分别>75%和>55%),中部相对低(15%~55%)的特征,与其土地利用/覆被特征相吻合。

关键词: 植被覆盖度, 小波能量系数, 主成分分析, 植被指数

Abstract:

The vegetation cover (VC) and corresponding vegetation canopy reflectance curves were collected in "Remotely sensed loading integrated testing site of non-driving aircraft (North testing site)"of Peking University, and the VC estimation models were developed and compared with each other based on the correlation between the conversion types (wavelet energy coefficient, principal component and vegetation index) of hyper-spectral curves and VC value. The result indicates: The hyper-spectral vegetation index-based model (R2>0.8, RMSE≤0.0188) is the best one of the three conversion types-based models, and EVI-based model is the best one among the other vegetation index-based models; the higher correlation coefficients existed between the second and the fourth single wavelet energy coefficient retrieved from 8-scale wavelet transformation and VC value (R2=0.811 and 0.813; RMSE=0.0199 and 0.0198, respectively); the multi-regression model established between multiple single wavelet energy coefficients and VC works better than the model based on the principal component analysis, the R2 and RMSE were improved by 0.058 and 0.03, respectively; the VC spatial distribution map through combining EVI-based model and TM-EVI indicates that the higher VC is distributed in the northern (75%) and southern (55%) parts of the study site and the lower VC (15%-55%) is distributed in the middle part. The spatial distribution is consistent with the land use/cover characteristics.

Key words: vegetation cover, wavelet energy coefficient, principal component analysis, vegetation index

中图分类号: 

  • TP79