摘要
以滇西北典型代表区域的迪庆州香格里拉市云冷杉为研究对象,运用Landsat 8遥感影像数据结合地面角控样地调查数据,建立云冷杉BPNN和SVM估测模型,进行对比分析。结果表明:SVM模型精度明显优于BPNN模型,其R2、r RMSE和P分别为0.67、27.91%和77.09%。利用SVM遥感估测模型得到香格里拉市云冷杉林总蓄积量与传统森林资源二类调查当年的统计结果误差仅为1.14%,SVM估测模型可为今后森林蓄积量估测提供参考。
Selected the typical forest types of spruce-fir forests in Shangri-La City, located in Diqing of Northwest Yunnan, as the research objects. Using Landsat 8 remote sensing image data combined with ground angle control plot survey data, the BPNN and SVM estimation models of spruce-fir were established and compared. The results show that the precision of the SVM model is obviously better than that of the BPNN model, its R2、rRMSE and P are 0.67, 27.91% and 77.09% respectively. The statistical error between the total volume of spruce-fir forests in Shangri-La City and the results of the second-class survey of traditional forest resources is only 1.14% by using SVM remote sensing estimation model, which shows that SVM estimation model could provide support for forest resources.
引文
[1]孟宪宇.测树学[M].北京:中国林业出版社,2006.
[2]李崇贵,赵宪文,李春干.森林蓄积量遥感估测理论与实现[M].北京:科学出版社,2006:1.
[3]何亚娟,潘学标,裴志远,等.基于SPOT遥感数据的甘蔗叶面积指数反演和产量估算[J].农业机械学报,2013,44(5):226-231.
[4]黄平,候长谋,杨燕琼.基于RS、GIS的杉木林分蓄积量判读模型研究[J].中南林业调查规划,2003,22(1):25-27.
[5]Pai P F,Hong W C.Support vector machines with simulated annealing algorithms inelectricity load forecasting[J].Energy Conversion and Management,2005,46(17):2669-2688.
[6]杨永恬.基于多源遥感数据的森林蓄积量估测方法研究[D].北京:中国林业科学研究院,2010.
[7]刘治国,王弟,张红,等.基于遥感数据的山西东部油松林蓄积量估测方法研究[J].山西大学学报(自然科学版),2017,40(1):187-194.
[8]李际平,邓立斌,何建华.基于人工神经网络的森林资源预测研究[J].中南林学院学报,2001,21(4):19:22.
[9]王臣立,牛铮,郭治兴,等.基于植被指数和神经络的热带人工林地上蓄枳遥感估测[J].生态环境学报,2009,18(5):1830-1834.
[10]岳彩荣.香格里拉县森林生物量遥感估测研究[D].北京:北京林业大学,2011.
[11]汉卿.森林蓄积量的支持向量机估测理论方法研究[D].西安:西安科技大学,2014.
[12]刘青.人工神经网络遥感影像分类系统的设计与实现[D].昆明:昆明理工大学,2010.
[13]Cherkassky V,Ma Y Q.Practical selection of SVMparameters and noise estimation for SVM regression[J].Neural Networks,2004,17(1):113-126.