首页|基于BP神经网络的落叶松树冠体积及表面积模型构建

基于BP神经网络的落叶松树冠体积及表面积模型构建

扫码查看
[目的]应用BP神经网络模型预测华北落叶松树冠体积与表面积,探索华北落叶松树冠体积与表面积估算模型的最优形式,为未来的预测模式提供新思路。[方法]以山西省庞泉沟自然保护区的华北落叶松林为研究对象,通过从6块(60 m×60 m)固定样地得到的678个观测数据,运用BP神经网络,分别对华北落叶松树冠体积与表面积建立模型,通过对模型的训练,得到基于BP神经网络的华北落叶松树冠体积和表面积估算模型。[结果]基于BP神经网络的华北落叶松树冠体积与表面积模型的最优结构模型的输入层节点数∶隐层节点数∶输出层节点数=6∶9∶1。其中树冠体积的决定系数(R2)、平均绝对误差(MAE)和均方根误差(RMSE)分别为0。948、5。40 m3、18。40;表面积的R2、MAE和RMSE分别为0。957、3。33 m2、14。41。基于BP神经网络的华北落叶松树冠体积与表面积模型的性能与输入因子的数量呈正相关,最优模型的输入因子数为6个,分别为冠幅、树高、胸径、最大冠幅高度、第一活枝长(在垂直于树干方向上的投影长度)和冠基高。[结论]输入变量包含树干尺寸和树冠构型特征相关信息,模型能较好地实现华北落叶松树冠体积和表面积的预测。
Model construction of Larix principis-rupprechtii canopy volume and surface area based on BP neural network
[Objective]The BP neural network model was applied to predict the canopy volume and surface area of Larix principis-rupprechtii,and the optimal form of canopy volume and surface area estimation model of L. principis-rupprechtii was explored in order to provide new ideas for the future prediction model.[Method]Taking L. principis-rupprechtii in Pangquangou Nature Reserve of Shanxi Province,northern China as the research object,the canopy volume and surface area of L. principis-rupprechtii were constructed using BP neural network based on 678 observational data obtained from six (60 m × 60 m) fixed plots.[Result]Through model training,the canopy volume and surface area estimation model of L. principis-rupprechtii was obtained based on BP neural network. Based on BP neural network,the number of input layer nodes∶number of hidden layer nodes∶number of output layer nodes was 6∶9∶1. Canopy volume data R2=0.948,MAE=5.40 m3,RMSE=18.40;surface area data R2=0.957,MAE=3.33 m2,RMSE=14.41. The performance of L. principis-rupprechtii canopy volume and surface area model based on BP neural network was positively correlated with the number of input factors. The optimal model had 6 input factors,i.e. crown width,tree height,DBH,max. crown height,projection length of the first live branch in the direction perpendicular to trunk,and crown base height.[Conclusion]The input variables include information related to trunk size and crown configuration characteristics. The model can realize the prediction for the crown volume and surface area of Larix principis-rupprechtii trees effectively.

model buildingscanopy volume and surface areaBP neural networkmachine learningprediction modelLarix principis-rupprechtii

周来、程小芳、张梦弢

展开 >

山西农业大学林学院,山西太谷 030801

模型构建 树冠体积与表面积 BP神经网络 机器学习 预测模型 华北落叶松

山西省高等学校科技创新项目北京林业大学中央高校基本科研业务费专项山西农业大学博士科研启动项目国家自然科学基金青年科学基金项目

2021L138BFUKF2023042021BQ1531901308

2024

北京林业大学学报
北京林业大学

北京林业大学学报

CSTPCD北大核心
影响因子:1.237
ISSN:1000-1522
年,卷(期):2024.46(8)