基于不同分析模型的大豆叶片SPAD值和LAI光谱估算比较  

Comparison of SPAD Value and LAI Spectral Estimation of Soybean Leaves Based on Different Analysis Models

作　　者：唐子竣 向友珍 王辛[1,2] 安嘉琪 郭金金 王晗 李志军 张富仓 TANG Zijun;XIANG You-zhen;WANG Xin;AN Jia-qi;GUO Jin-jin;WANG Han;LI Zhi-Jun;ZHANG Fu-cang(Institute of Water-saving Agriculture in Arid Areas of China/Key Laboratory of Agricultural Soil and Water Engineering in Arid and Semiarid Areas of Ministry of Education,Northwest A&F University,Yangling 712100,China;College of Water Resources and Architectural Engineering,Northwest A&F University,Yangling 712100,China)

机构地区：[1]西北农林科技大学旱区节水农业研究院/旱区农业水土工程教育部重点实验室,陕西杨凌712100 [2]西北农林科技大学水利与建筑工程学院,陕西杨凌712100

出　　处：《大豆科学》2023年第1期55-63,共9页Soybean Science

基　　金：国家自然科学基金(52179045)。

摘　　要：为探讨在大豆鼓粒期采用光谱技术估算叶片SPAD值和LAI的有效分析模型和方法,本研究以大田鼓粒期大豆为试验材料,在3个不同时段9:45~10:15(10AM)、11:45~12:15(12PM)和13:45~14:15(2PM)测量冠层全波段光谱反射率,并分别使用极限学习机(ELM)、偏最小二乘回归(PLSR)、支持向量机(SVM)和随机森林(RF)构建大豆叶片SPAD值和LAI估算模型,并对比不同模型分析结果的估算精度。结果表明:在各模型中,12PM和2PM测定的光谱反射率与大豆叶片SPAD值和LAI的拟合精度均高于10AM。基于RF的大豆叶片SPAD值估算模型验证集的R2为0.910,RMSE为2.006,MRE为3.684;基于RF的大豆LAI估算模型验证集的R^(2)为0.916,RMSE为0.209,MRE为4.383,与ELM、PLSR和SVM相比,有更高的估算精度。综上结果说明大豆鼓粒期在11:45~12:15和13:45~14:15采用RF模型,运用全波段的光谱反射率估算大豆叶片SPAD值和LAI可得到较准确的结果。In order to explore the effective analytical model and methods for estimating leaf SPAD value and LAI by spectroscopic technique in soybean drum-grain stage, this study used soybean at seed-filling stage in the field as the test material, conducted three different time periods at 9:45-10:15(10 AM), 11:45-12:15(12 PM) and 13:45-14:15(2 PM) to measure the canopy full-band spectral reflectance, used extreme learning machine(ELM), partial least squares regression(PLSR), support vector machine(SVM) and random forest(RF) to build soybean leaf SPAD value and LAI estimation models, and compared the estimation accuracy of the analysis results of different models. The results showed that in each model, the fitting accuracy of the spectral reflectance measured by 12 PM and 2 PM with the SPAD value and LAI of soybean leaves was higher than that of 10 AM. The R^(2) of the RF-based soybean leaf SPAD value estimation model validation set was 0.910, the RMSE was 2.006, and the MRE was 3.684. The RF-based soybean LAI estimation model validation set was with R^(2) of 0.916, RMSE was 0.209, MRE was 4.383, compared with ELM, PLSR and SVM, it had higher estimation accuracy. The results also showed that the full-band spectral reflectance took on at 11:45-12:15 and 13:45-14:15 with the RF model could be used to estimate the SPAD value and LAI of soybean leaves, which could obtain more accurate results.

关 键 词：随机森林 光谱技术 大豆 鼓粒期 SPAD值 叶面积指数 

分 类 号：S565.1[农业科学—作物学] S127