武旭梅,常庆瑞,落莉莉,由明明.水稻冠层叶绿素含量高光谱估算模型[J].干旱地区农业研究,2019,37(3):238~243
水稻冠层叶绿素含量高光谱估算模型
Hyperspectral estimation model for chlorophyll content of rice canopy
  
DOI:10.7606/j.issn.1000-7601.2019.03.31
中文关键词:  高光谱遥感  估算模型  随机森林算法  水稻冠层  叶绿素  植被指数
英文关键词:hyperspectral remote sensing  estimation model  random forest algorithm  rice canopy  chlorophyll content  vegetation index
基金项目:国家863计划项目(2013AA102401)
作者单位
武旭梅 西北农林科技大学资源环境学院陕西 杨凌 712100 
常庆瑞 西北农林科技大学资源环境学院陕西 杨凌 712100 
落莉莉 西北农林科技大学资源环境学院陕西 杨凌 712100 
由明明 西北农林科技大学资源环境学院陕西 杨凌 712100 
摘要点击次数: 845
全文下载次数: 447
中文摘要:
      为了寻求西北引黄灌区水稻冠层叶绿素含量的高精度估算模型,通过田间试验测定了水稻冠层SPAD和高光谱数据,运用任意波段组合的方式构建了一系列基于原始光谱、一阶导数光谱的比值、差值、归一化和土壤调节植被指数,筛选出反映水稻冠层SPAD的最佳植被指数作为自变量,应用普通回归分析方法和随机森林算法建立了该区域水稻冠层SPAD估算模型并进行了对比分析。结果表明:(1)应用普通回归分析方法,以RVI(D1316,D736)为自变量建立的指数模型是估算西北引黄灌区水稻冠层SPAD的最佳单变量模型;(2)采用随机森林算法,以4个植被指数RVI(R696,R540)、DVI(R700,R536)、SAVI(R700,R536)、RVI(D1316,D736)建立的估算模型比普通回归模型精度更高,验证结果的决定系数R2为0.873,均方根误差RMSE为3.221,平均相对误差RE为13.25%。说明通过随机森林算法建立的模型可以实现水稻冠层SPAD的精准估测,可以用于西北引黄灌区水稻冠层叶绿素含量的快速、无损获取。
英文摘要:
      In order to establish a high-precision model to estimate rice canopy chlorophyll content in Northwest Yellow River Irrigation Area, rice canopy SPAD and hyperspectral data were measured in a field experiment. In this paper, a series of Ratio Vegetation Index(RVI), Difference Vegetation Index(DVI), Normalized Difference Vegetation Index (NDVI), and Soil-Adjust Vegetation Index (SAVI) were computed by the combination of original canopy spectra. The optimal vegetation indexes that sensitively reflected the rice canopy SPAD were screened out. The estimation models of rice canopy SPAD in Northwest Yellow River Irrigation Area were established by using ordinary regression analysis method and random forest algorithm. The results showed that: (1) Using the general regression analysis method, the exponential model established with RVI (D1316, D736) as the independent variable was the best single-variable model to estimate rice canopy SPAD in the study area. (2) Compared with the normal regression models, the random forest model established using four vegetation indices RVI (R696, R540), DVI (R700, R536), SAVI (R700, R536), and RVI (D1316, D736) had the best prediction accuracy. The validation results showed that R2 was 0.873, RMSE was 3.221, and RE was 13.25%. Therefore, we concluded that the random forest model can be used for the rapid and lossless estimation of rice canopy chlorophyll content in Northwest Yellow River Irrigation Area.
查看全文  查看/发表评论  下载PDF阅读器