当前位置: 首页 > 文章 > 基于SVM方法的高山松林蓄积量遥感估测研究 西部林业科学 2014 (4) 83-88
Position: Home > Articles > Study on the Remote Sensing Estimation of Pinus densata's Stock Volume Based on SVM Journal of West China Forestry Science 2014 (4) 83-88

基于SVM方法的高山松林蓄积量遥感估测研究

作  者:
付虎艳;徐云栋;李圣娇;苏院兴;舒清态
单  位:
西南林业大学
关键词:
香格里拉县;高山松;蓄积量估测;SVM
摘  要:
以香格里拉县高山松为研究对象,利用2006年香格里拉县TM遥感影像、2006年森林资源二类调查小班数据、2009年精度为30 m的DEM数据以及2013年香格里拉县高山松实测样地数据,提取研究区内高山松林影像分布图及筛选出17个因子(13个遥感因子、3个地形因子、1个地面调查因子)作为备选自变量,在MATLAB下利用LIBSVM模块建立研究区高山松林蓄积量单位面积(30 m×30 m)估测模型。结果表明,选用RBF核函数在参数范围内寻找出SVM模型的最佳参数C=3.580 9,g=0.1、p=0.01,利用最佳寻优参数建立SVM非参数模型,对SVM模型进行测试得到,均方根误差MSE=0.008 7,复相关系数R=0.51,相对误差RE=23.4%,估测精度为76.6%。以像元为单位,分块提取高山松林对应的各像元自变量因子,利用估测模型预测得到香格里拉县高山松林总蓄积量为13 318 476.5 m3。
译  名:
Study on the Remote Sensing Estimation of Pinus densata's Stock Volume Based on SVM
作  者:
FU Hu-yan;XU Yun-dong;LI Sheng-jiao;SU Yuan-xing;SHU Qing-tai;Southwest Forestry University;
关键词:
Shangri-La County;;Pinus densata;;stock volume estimation;;SVM
摘  要:
By taking the Pinus densata in Shangri-La County as the research target,TM remote sensing image of Shangri-La County in 2006,the forest resources inventory data in 2006,the DEM data of 30 meters precision in2009,and the Pinus densata's ground sample data of Shangri-La County in 2013 was adopted as the data source.The Pinus densata's distribution image in the study area was extracted,and 17 factors( 13 remote sensing factors,3 terrain factors,1 ground survey factors) was selected as the alternative variables. By using LIBSVM module in MATLAB,the estimation model of Pinus densata's per unit( 30 m × 30 m) stock volume of study area was established. The results showed that the best optimal parameters were C = 3. 580 9,g = 0. 1,p = 0. 01 by using the RBF kernel function in the range of constant,MSE = 0. 008 7,R = 0. 51,RE = 23. 4 %,in SVM model test with estimation accuracy of 76. 6 %,and the predicted total volume of Pinus densata was 13 318 476. 5 m3 in Shangri-La county by taking the pixel as unit and extracting the independent variable factors.

相似文章

计量
文章访问数: 9
HTML全文浏览量: 0
PDF下载量: 1

所属期刊

推荐期刊