贾洁琼,刘万青,孟庆岩,孙云晓,孙震辉(西北大学, 城市与环境学院, 西安 710127;中国科学院遥感与数字地球研究所, 北京 100101;三亚中科遥感研究所, 三亚 572029)
目的 叶面积指数（LAI）是重要的植被生物理化参数，对农作物长势和产量预测具有重要研究意义。基于物理模型和经验模型的LAI估算方法被认为是当前最常用的方法，但两种方法的估算效率和精度有限。近年来，机器学习算法在遥感监测领域广泛应用，算法具有描述非线性数据拟合、融合更多辅助信息的能力，为了评价机器学习算法在玉米LAI遥感估算中的适用性，本文分析比较了随机森林和BP神经网络算法估算玉米LAI的能力，并与传统经验模型进行了比较。方法 以河北省怀来县东花园镇为研究区，基于野外实测玉米LAI数据，结合同时期国产高分卫星（GF1-WFV影像），首先分析了8种植被指数与LAI的相关性，进而采用保留交叉验证的方式将所有样本数据分为两部分，65%的数据作为模型训练集，35%作为验证集，重复随机分为3组，构建以8种植被指数为自变量，对应LAI值为因变量的RF模型、BP神经网络模型及传统经验模型。采用决定系数R2和均方根误差（RMSE）作为模型评价指标。结果 8种植被指数与LAI的相关性分析表明所有样本数据中，实测LAI值与各植被指数均在（P<0.01）水平下极显著相关，且相关系数均高于0.5；将3组不同样本数据在随机森林、BP神经网络算法中多次训练，并基于验证数据集进行估算精度检验，经验模型采用训练数据集建模，验证数据集检验，结果表明，RF模型表现出了较强的预测能力，LAI预测值与实测值R2分别为0.681、0.757、0.701，均高于BP模型（0.504、0.589、0.605）和经验模型（0.492、0.557、0.531），对应RMSE分别为0.264、0.292、0.259；均低于BP模型（0.284、0.410、0.283）和经验模型（0.541、0.398、0.306）。结论 研究表明，RF算法能更好地进行玉米LAI遥感估算，为快速准确进行农作物LAI遥感监测提供了技术参考。
Estimation of maize leaf area index based on GF-1 WFV image and machine learning random algorithm
Jia Jieqiong,Liu Wanqing,Meng Qingyan,Sun Yunxiao,Sun Zhenhui(College of Urban and Environmental Sciences, Northwest University, Xi'an 710127, China;Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China;Institute of Remote Sensing of Sanya Sanya 572029, China)
Objective Leaf area index (LAI) is an important biological and physical parameter of vegetation, and it plays an important role in predicting crop growth and yield. A number of LAI estimation methods have been developed from remotely sensed data, each of which presents unique advantages and limitations. The empirical regression and physical models are the most widely used among these methods. The empirical regression model can reduce the effect of background noise on the spectral reflectance of plant canopies, and the physical model simulates the radiative transfer process in vegetation and describes the canopy spectral variation as a function of canopy, leaf, and soil background characteristics. However, the efficiency and accuracy of the two methods are limited. In recent years, machine learning algorithms have been widely used in remote sensing monitoring, and they can describe nonlinear data fitting and fuse more auxiliary information. This study evaluates the applicability of machine learning algorithms in maize LAI remote sensing estimation. Method In this study, the east garden of Huailai County in Hebei Province was used as the study area. Eight kinds of vegetation indices based on the GF1 WFV satellite images were calculated, and the correlation between the same-period measured LAI and the vegetation index was analyzed. Then, all the in situ measured corn LAI and corresponding eight vegetation indices were randomly divided into a training dataset and an independent model validation dataset (65% and 35% of the data, respectively). These datasets were randomly divided into three groups repeated three times. The training dataset was used to establish models to predict corn LAI, and the validation dataset was employed to test the quality of each prediction model. Finally, utilizing random forest, backpropagation (BP) neural network algorithm, and the traditional empirical model, the LAI inverting model was established based on previous work. This study compared the estimation accuracy of the three models for each sample group on the basis of the coefficient of determination (R2) and root mean square error (RMSE) to evaluate the estimation accuracy of each model and to compare the performances of the three models further. Result Results showed that the LAI values were significantly correlated with the vegetation index at the P < 0.01 level in all the sample data and that the correlation coefficients were higher than 0.5. Three groups of different sample data were trained in random forest and BP neural network for many times, and the accuracy of estimation was checked based on the validation dataset. The empirical model was established by training dataset and verified by validation dataset. The results show that the RF model outperformed BP and the traditional empirical model in each group of sample data. For the RF models, R2 of the estimated and measured LAI values were 0.681, 0.757, and 0.701 in contrast to the RMSE of 0.264, 0.292, and 0.259, respectively. For the BP model, R2 for the three groups was 0.504, 0.589, and 0.605, and the corresponding RMSE was 0.284, 0.410, and 0.283, respectively. However, for the traditional empirical model, R2 for the three groups was 0.492, 0.557, and 0.531, and the corresponding RMSE was 0.541, 0.398, and 0.306, respectively. Conclusion The RF algorithm provides an effective approach to improve the prediction accuracy of corn LAI and provides a technical reference for the rapid and accurate monitoring of crop LAI remote sensing.