摘要
叶面积指数(LAI)是反映马尾松生长状况的一个重要参数,快速、准确、无损地估测马尾松LAI能为马尾松的经营管理提供基础数据。使用LAI-2200型植物冠层分析仪获取花溪区马尾松样地LAI数据,结合同期Landsat 8 OLI数据,选择并计算了与LAI密切相关的8种植被指数,分析了各种植被指数与样地实测LAI的相关性,进而使用天牛须搜索(BAS)优化的BP神经网络模型构建了马尾松LAI遥感估算模型,以反向传播神经网络(BP)模型、遗传算法(GA)优化的BP神经网络模型和粒子群(PSO)优化的BP神经网络为参比模型,以决定系数(R^(2))、均方根误差(RMSE)和CPU运行时间为指标评价并比较了模型估算精度。结果表明:全样本数据中,各植被指数均与对应的LAI呈现极显著相关(P<0.01),相关系数都大于0.5;BAS-BP模型在3个样本组中的预测精度和训练速度均高于同期的BP模型、GA-BP模型和PSO-BP模型;3个样本组中BAS-BP模型的LAI预测值与实测值的R^(2)分别为0.6624、0.6949和0.7163,均高于同期的BP模型、GA-BP模型和PSO-BP模型,对应的RMSE分别为0.4181、0.3759和0.3798,训练时间分别为44.24、42.08 s和41.72 s,均小于同期的3种模型。因此,BAS-BP可作为快速、准确估算马尾松LAI的一种新方法。
Leaf Area Index(LAI)is an important parameter to reflect the growth status of Pinus massoniana.Rapid,accurate and non-destructively estimation of LAI can provide basic data for the management of Pinus massoniana.LAI-2200 plant canopy analyzer was used to obtain the LAI data of Pinus massoniana sample plot in Huaxi District.Totally 8 kinds of vegetation index based on the Landsat 8 OLI image in the same period were calculated and the correlation between measured LAI and the vegetation index was analyzed.The purpose was to construct Pinus massoniana LAI estimation model specifically by using the back-propagation neural network model(BP)based on the optimization of Beetle Antennae Search(BAS).Additionally for each sample group,the models based on BP model,the Genetic Algorithm(GA)BP model and the Particle Swarm Optimization(PSO)BP model were employed as comparison models.The estimation accuracy of the model was evaluated and compared with the determination coefficient(R^(2)),Root Mean Square Error(RMSE)and CPU running time.The results indicated that:in the whole sample data,each vegetation index has a significant correlation with the corresponding LAI.The correlation coefficient is greater than 0.5;the prediction accuracy and training speed of BAS-BP model in the three sample groups are higher than that of BP model,GA-BP model and PSO-BP model in the same period;the LAI prediction value of BAS-BP model and the measured value of R^(2) in the three sample groups are 0.6624,0.6949 and 0.7163,respectively,which are high.The corresponding RMSE of BP model,GA-BP model and PSO-BP model in the same period are 0.4181,0.3759 and 0.3798 respectively,and the training time is 44.24s,42.08s and 41.72s respectively,which are less than the other three models.Therefore,BAS-BP can be used as a new method to estimate LAI of Pinus massoniana.
作者
毕陈权
石振情
谭伟
朱玉婷
周浩
程旺
Bi Chenquan;Shi Zhenqing;Tan Wei;Zhu Yuting;Zhou Hao;Cheng Wang(Guizhou Technological College of Ecology and Energy,Guiyang 550001,Guizhou,China;College of Forestry,Guizhou University,Guiyang 550025,Guizhou,China;Research Center of Forestry Information Engineering,Guizhou University,Guiyang 550025,Guizhou,China)
出处
《绿色科技》
2025年第2期53-60,共8页
Journal of Green Science and Technology
基金
贵州省科技支撑计划项目(编号:黔科合支撑[2017]2520-1号)。
关键词
马尾松
叶面积指数
遥感估算
天牛须搜索(BAS)算法
BP神经网络
Pinus massoniana
leaf area index
remote sensing estimation
Beetle Antennae Search(BAS)algorithm
BP neural network
