基于梯度提升树算法的夏玉米叶面积指数反演  

Inversion of Summer Maize Leaf Area Index Based on Gradient Boosting Decision Tree Algorithm

作　　者：张宏鸣[1] 刘雯 韩文霆[2] 刘全中[1] 宋荣杰[1] 侯贵河 ZHANG Hongming;LIU Wen;HAN Wenting;LIU Quanzhong;SONG Rongjie;HOU Guihe(College of Information Engineering,Northwest A&F University,Yangling,Shaanxi 712100,China;College of Mechanical and Electronic Engineering,Northwest A&F University,Yangling,Shaanxi 712100,China)

机构地区：[1]西北农林科技大学信息工程学院,陕西杨凌712100 [2]西北农林科技大学机械与电子工程学院,陕西杨凌712100

出　　处：《农业机械学报》2019年第5期251-259,共9页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家重点研发计划项目(2017YFC0403203);国家自然科学基金项目(41771315;41301283;41371274);欧盟地平线2020研究与创新计划项目(GA:635750)

摘　　要：为了快速、准确、大范围获取大田夏玉米的叶面积指数(Leaf area index,LAI),基于实地采集的夏玉米LAI和株高,结合同时期的无人机多光谱影像,选择与夏玉米LAI相关性较强的8种植被指数以及株高作为反演LAI的输入变量,采用梯度提升树(Gradient boosting decision tree,GBDT)算法建立植被指数及株高与叶面积指数之间的预测模型,并与支持向量机(Support vector machine,SVM)和随机森林(Random forest,RF)算法建立的模型进行预测精度对比。结果表明,GBDT算法在3个样本组中的LAI预测值与实测值R^2分别为0. 571 0、0. 755 8、0. 644 1,均高于SVM算法(0. 547 2、0. 679 1、0. 616 8)和RF算法(0. 550 5、0. 697 3、0. 629 5);对应的RMSE分别为0. 002 7、0. 001 5、0. 001 6,均低于SVM算法(0. 211 7、0. 152 3、0. 159 7)和RF算法(0. 244 7、0. 214 7、0. 208 0)。该研究为快速准确的大田夏玉米LAI遥感监测提供了技术和方法。Aiming to obtain the leaf area index (LAI) of large-scale summer maize in a high-efficiency, non-destructive and large-scale manner, and provide a technical reference for remote sensing monitoring of summer maize growth. The research was based on the field-collected summer maize LAI and maize height, as well as combined with multi-spectral data of the same period, eight vegetation indexes and height with strong correlation with summer maize LAI were selected as the input variables of gradient boosting decision tree (GBDT) algorithm model for LAI inversion. The support vector machine (SVM) model and the random forest (RF) model were taken as the reference models, which were used to compare the accuracy of prediction. The results showed that the GBDT algorithm model prediction consequence were better than the other two models among the three sample groups. R^2 of prediction value and measured LAI values of the sample groups 1, 2 and 3 were 0.571 0, 0.755 8 and 0.644 1, respectively, which were higher than those of the SVM models (0.547 2, 0.679 1, 0.616 8) and RF models (0.550 5, 0.697 3, 0.629 5), corresponding root mean square error (RMSE) values were 0.002 7, 0.001 5 and 0.001 6, which were lower than those of the SVM model (0.211 7, 0.152 3 and 0.159 7) and RF model (0.244 7, 0.214 7 and 0.208 0). The research result provided a technical method for fast and accurate monitoring of summer maize LAI remote sensing in the field.

关 键 词：夏玉米 叶面积指数 多光谱 梯度提升树 植被指数 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置] S127[自动化与计算机技术—控制科学与工程]