摘要
基于Landsat TM影像和DEM数据,尝试利用BP神经网络建立旺业甸林场森林生物量非线性遥感模型系统,通过实验筛选,最终利用增强型的BP网络进行训练仿真。模型仿真结果表明,增强型的BP神经网络具有自学习和自适应功能强、收敛速度快的特点,能够最大限度地利用先验样本。仿真检验结果的相对系数达0.802 2,平均相对误差为15.7%,表明该模型预测的生物量与实际生物量一致性较好,能够达到较好的反演效果。
Based on Landsat TM images, DEM data, and using BP neural network, this study established Wangyedian forest farm biomass non-linear remote sensing model system. Choosing by experiment and taking enhancement mode of BP network, simulation training has been conducted. The results showed that： the enhancement model of BP neural network has characteristics of self-learning, strong adaptive and fast convergence . It could maximum use prior sample, and relative coefficient reached to 0. 802 2, the average relative error was 15.7%. This model also had great consistency of predicted biomass and actual biomass, could achieve better inversion effect.
出处
《林业调查规划》
2013年第1期5-8,共4页
Forest Inventory and Planning
