芦苇地上生物量的SDBP神经网络估算模型研究

STUDY ON SDBP NEURAL NETWORK ESTIMATION MODEL ABOUAT BIOMASS OF REED ON THE GROUND

为了快速准确的测量芦苇的生物量,设计了SDBP神经网络模型。模型的输入参数为芦苇的植株平均高度和植株密度,输出参数为芦苇地上生物量,利用60组数据经过训练后,其测试样本的网络输出与目标函数的相关系数达到0.995 91,用训练好的网络模型对15组未参加训练的数据进行生物量测定,其测量结果与实际结果的标准差较小(最大标准差为1.091 2),可以满足实际测量需要。训练好的神经网络模型可以在不破坏芦苇植株的前提下准确、快速、大面积测量芦苇生物量。

in order to quickly and accurately measure biomass of reed, the SDBP neural network model is designed. Inputs of the model are average height and density of plants, and outputs are biomass of reed on the ground. After training by using 60 sets of data, the correlation function of network output of test sample and objective function is 0.99591. 15 sets of data are used to verify the trained model, and Standard deviation of the measurements and actual results is smaller （ the maximum error is 1. 0912）, which satisfy the requirement of actual measurement. The trained model can measure quickly and accurately biomass of reed without damaging plants.