论文部分内容阅读
选择浙江省内临安、安吉、龙泉3个毛竹产区为研究区域,基于野外调查数据和Landsat 5 TM影像,分别建立3个区域的毛竹林生物量遥感估算模型,包括一元线性模型、一元非线性模型、逐步回归模型、多元线性模型和Erf-BP神经网络模型,并对3个区域的模型进行评价;最后,选择精度较好的模型进行移植并对其可移植性进行分析.结果表明:在3个区域,Erf-BP神经网络模型精度均最高,逐步回归模型和一元非线性模型次之.Erf-BP神经网络模型的可移植性最佳.模型类型和模型自变量对统计模型的可移植性有较大影响.
Based on the field survey data and Landsat 5 TM images, three remote sensing models of bamboo biomass were established in Lin’an, Anji, Longquan, Zhejiang Province, including the unilinear linear model, unary nonlinear model Model, stepwise regression model, multivariate linear model and Erf-BP neural network model, and evaluated the model of the three regions.Finally, the model with better accuracy was selected for transplantation and its portability was analyzed.The results showed that in the The Erf-BP neural network model has the highest accuracy in all three regions, followed by the stepwise regression model and the unary nonlinear model.The Erf-BP neural network model has the best portability.The portability of model type and model independent variable to the statistical model Sex has a greater impact.