落莉莉,常庆瑞,武旭梅,杨景,李粉玲,王琦.夏玉米叶片光合色素含量高光谱估算[J].干旱地区农业研究,2019,37(4):178~183
夏玉米叶片光合色素含量高光谱估算
Hyperspectral estimation of photosynthetic pigment contents of summer maize leaves
  
DOI:10.7606/j.issn.1000-7601.2019.04.24
中文关键词:  夏玉米  光合色素  高光谱估算  随机森林算法  关中地区
英文关键词:summer maize  photosynthetic pigments  hyperspectral estimation  random forests  Guanzhong district
基金项目:国家高技术研究发展计划(863计划)项目(2013AA102401-2);中央高校基本科研业务费专项资金项目(2452017108)
作者单位
落莉莉 College of Nature Resources and Environment, Northwest A&F University,Yangling, Shaanxi 712100, China 
常庆瑞 College of Nature Resources and Environment, Northwest A&F University,Yangling, Shaanxi 712100, China 
武旭梅 College of Nature Resources and Environment, Northwest A&F University,Yangling, Shaanxi 712100, China 
杨景 College of Nature Resources and Environment, Northwest A&F University,Yangling, Shaanxi 712100, China 
李粉玲 College of Nature Resources and Environment, Northwest A&F University,Yangling, Shaanxi 712100, China 
王琦 College of Nature Resources and Environment, Northwest A&F University,Yangling, Shaanxi 712100, China 
摘要点击次数: 34
全文下载次数: 39
中文摘要:
      为了实现夏玉米叶片光合色素含量的快速、无损检测,以陕西省关中地区夏玉米“大丰26号”为研究对象,探究了不同总色素含量水平的玉米叶片反射光谱特征。分别提取与叶绿素a、叶绿素b、类胡萝卜素和总色素含量相关性较强的15个光谱参数,通过单变量回归、多元逐步回归和随机森林回归分析,建立光合色素含量估算模型并进行精度比较。结果表明:基于随机森林方法构建的光合色素估算模型精度最高,其中,叶绿素a、叶绿素b、类胡萝卜素的建模R2为0.93,总色素的建模R2为0.92;叶绿素a和类胡萝卜素的检验R2<、sup>为0.74,叶绿素b和总色素的检验R2为0.71;各模型的均方根误差(RMSE)和相对误差(RE)相差不大;拟合精度由高到低依次为叶绿素a、类胡萝卜素、总色素和叶绿素b的RF模型。证实了随机森林方法在夏玉米叶片光合色素含量估算中的优越性,并构建了高精度的光合色素RF估算模型。
英文摘要:
      To realizing the rapid and non-destructive detection of photosynthetic pigment content of summer maize, the spectral reflectance characteristics of maize leaf with different total pigment contents were studied by taking “Dafeng 26” of summer maize in Guanzhong district of Shaanxi Province as the research object. First, 15 spectral parameters with good correlation of chlorophyll a, chlorophyll b, carotenoids and total pigments are extracted. Then, the estimation models of photosynthetic pigments were established by single variable regression, multiple stepwise regression and random forest regression analysis, respectively. The results showed that the estimation model of photosynthetic pigments based on random forest had the highest accuracy. Specifically, the R2 of fitting model of chlorophyll a, chlorophyll b and carotenoids was 0.93, the R2 of fitting model of total pigments was 0.92, the R2 of validation model of chlorophyll a and carotenoids was 0.74, and the R2 of validation model of chlorophyll b and total pigments was 0.71. The RMSE and RE of each model are similar. The fitting accuracy from high to low was the RF model of chlorophyll a, carotenoids, total pigments and chlorophyll b, respectively. The study constructed a high-precision RF estimation model of photosynthetic pigments, which has application and population value.
查看全文  查看/发表评论  下载PDF阅读器