基于RS和GIS的径向基神经网络模型对森林蓄积量的估测

Volume Estimation by Radial Basis Function Neural Network Model Based on RS and GIS Technologies

以塞罕坝机械林场的华北落叶松林为研究对象,利用SPOT5影像,基于RS和GIS确定蓄积量主要影响因子,即海拔、坡向、郁闭度、SP1、SP3、SPl/2、SP1-2/1+2、SP2*3/1,选取径向基神经网络模型中的广义回归神经网络模型对其蓄积量进行估测。结果表明:对林分蓄积量估测的最高精度为98.70%,最低精度为68.56%,预估检验的所有样地的平均精度为87.24%。利用径向基GRNN模型建立森林蓄积量估测模型对蓄积量进行估测时,效率高,计算方法比较简洁,易于操作。

Volume estimation by radial basis function neutral network model based on RS and GIS was conducted on Larix prmcipis-rupprechtii forest in Saihanba Jixie Forest Farm, Hebei province. Estimation was made on SPOT5 images by generalized regression of neural network in radial basis function neural network model, selecting main impact factors such as altitude, slope aspect, canopy density, SP1, SP3, SP1/2, SP1-2/1＋2, SP2＊3/1. The result showed that the highest precision of estimation of volume was 98,70% and the lowest one was 68.56%, averaging accuracy of 87.24%. It concluded that GRNN model had advantages of high efficiency, easy to calculate and operate for estimating forest volume.