马尾松人工林直径分布神经网络模型研究

该文首先以相对直径为输入变量，以累积频率为输出变量，构建了1∶S∶1的林分直径分布BP神经网络模型. 用1块具有代表性、26年生、全林伐倒测定每木胸径、树高生长过程的马尾松人工林标准地直径分布数据，作为马尾松人工林的期望分布，对所建模型进行训练、用定性与定量相结合的方法，选出既符合林分直径分布规律，又具有较高拟合准确度的网络模型结构为1∶2∶1，网络对象名为FRdnet2 该模型的总体累积频率拟合准确度达99.5%，径阶累积频率拟合准确度最低93%、最高99.9%、平均99%，径阶频率拟合准确度最低82%、最高99%、平均95%. 神经网络建模技术的拟合准确度好，可作为有效的林分直径分布模拟技术. 

Neural network models of diameter distribution for Pinus massoniana plantations

A BP neural network model of diameter distribution was created by using relative diameters of trees as input variables, and using accumulated frequencies of tree numbers as output variables, in which the model structure was 1∶S∶1 The expected value distribution of the training data of the neural network model are breast height diameter and diameter distribution, which are collected from a typical 26-year-old Pinus massoniana plantation plot and it is still in growth period. By using the combination of qualitative and quantitative methods, a network model is selected and it is correspond to the discipline of stand diameter distribution and has a high fitting accuracy. The network structure is 1∶2∶1 and named FRdnet2 The total accumulated frequency fitting accuracy of the model is 99.5%.Concretely, the accumulated frequency fitting accuracy of diameter class is 93% to 99.9% and the mean value is 99%. While the frequency fitting the accuracy of diameter class is 82% to 99% and the mean value is 95%. The degree of accuracy of the neural network model is very high, and the neural network modeling technology can be applied as an effective modeling technology for diameter distribution of trees.