神经网络在木材生长轮密度预测中的应用研究

摘要

研究木材生长轮密度准确预测问题,树林在生长过程中受到生态因子、培育措施、立地条件以及树木本身遗传等因素影响,木材生长轮密度呈非线性变化规律.针对传统线性预测方法只能对线性变化规律进行预测的缺陷,提出一种神经网络的木材生长轮密度预测方法.首先对木材生长轮密度一维数据进行重构变成多维数据,然后将数据输入到神经网络进行学习,并采用粒子群算法对神经网络参数进行优化,最后建立木材生长轮密度最优预测模型.采用具体木材生长轮数据对建立的最优预测模型性能进行仿真,结果表明,相对线性预测方法,改进方法提高木材生长轮密度预测精度,减少了预测误差,能刻画木材生长轮密度的变化趋势,是一种有效的木材生长密度变化的预测方法.%Research wood growth ring density prediction problem. Woods in the growth process are influenced by ecological factors, cultivation measures, site conditions and tree genetic and other factors, the density of wood growth ring changes nonlinearly. The paper proposed a wood growth ring density prediction method based on the neural network. The one - dimensional data of density of wood growth ring were reconstructed into.a multi - dimensional data which were input to the neural network for learning. The particle swarm optimization algorithm was used for the evolutionary of neural network parameters, Finally, the optimal prediction model of wood growth ring density was established. Using specific wood growth ring data to establish the optimal predicting model for performance testing, the results show that, compared with linear prediction method, this method can improve the prediction accuracy of density of wood growth ring, reduce the prediction error, and describe the change tendency of wood growth ring density, It is an effective wood density variation prediction method.