一种后向迭代的森林生物量遥感特征选择方法
发布时间：2018-08-18 21:03
【摘要】：森林生物量的估算对于全球碳平衡和环境保护至关重要。通过遥感等手段提取与森林生物量相关的单波段特征、植被指数、纹理特征、地形因子等特征参数,特征数量往往较多,影响预测精度。该文提出了一种后向迭代的随机森林(RF-RFE)特征选择方法,即利用随机森林算法计算特征重要度,采用后向迭代的方法逐步简化特征参数。以内蒙古大兴安岭地区的激流河林场为研究区域,以实验区2012年"资源三号"遥感影像和森林资源3类调查的样地数据为数据源,使用RF-RFE算法进行特征选择分析。实验结果表明,在森林生物量遥感反演过程中的RF-RFE特征选择不但能降低时间复杂度,而且保证了特征选择的精度。
[Abstract]:Estimation of forest biomass is critical to global carbon balance and environmental protection. The single band feature, vegetation index, texture feature, terrain factor and other characteristic parameters related to forest biomass are extracted by remote sensing, and the number of features is often large, which affects the prediction accuracy. In this paper, a backward iterative stochastic forest (RF-RFE) feature selection method is proposed, in which the feature importance is calculated by the stochastic forest algorithm, and the feature parameters are simplified step by the backward iterative method. Taking the Jiluhe forest farm in Daxinganling area of Inner Mongolia as the research area, and taking the remote sensing image of "resource 3" in the experimental area in 2012 and the sample data of forest resources investigation as data sources, the feature selection analysis was carried out by using RF-RFE algorithm. The experimental results show that RF-RFE feature selection can not only reduce the time complexity but also ensure the accuracy of feature selection.
【作者单位】： 山东农业大学信息科学与工程学院;中国测绘科学研究院;
【基金】：国家“863计划”项目(2013AA122003) 国家质量监督检验检疫总局公益性行业科研专项(201410308)
【分类号】：S771.8;TP751
本文编号：2190670
[Abstract]:Estimation of forest biomass is critical to global carbon balance and environmental protection. The single band feature, vegetation index, texture feature, terrain factor and other characteristic parameters related to forest biomass are extracted by remote sensing, and the number of features is often large, which affects the prediction accuracy. In this paper, a backward iterative stochastic forest (RF-RFE) feature selection method is proposed, in which the feature importance is calculated by the stochastic forest algorithm, and the feature parameters are simplified step by the backward iterative method. Taking the Jiluhe forest farm in Daxinganling area of Inner Mongolia as the research area, and taking the remote sensing image of "resource 3" in the experimental area in 2012 and the sample data of forest resources investigation as data sources, the feature selection analysis was carried out by using RF-RFE algorithm. The experimental results show that RF-RFE feature selection can not only reduce the time complexity but also ensure the accuracy of feature selection.
【作者单位】： 山东农业大学信息科学与工程学院;中国测绘科学研究院;
【基金】：国家“863计划”项目(2013AA122003) 国家质量监督检验检疫总局公益性行业科研专项(201410308)
【分类号】：S771.8;TP751
【相似文献】
相关期刊论文 前2条
1 徐新良;曹明奎;;森林生物量遥感估算与应用分析[J];地球信息科学;2006年04期
2 ;[J];;年期
相关会议论文 前1条
1 何红艳;郭志华;赵宪文;;遥感在森林地上生物量估算中的现状及展望[A];第十五届全国遥感技术学术交流会论文摘要集[C];2005年
相关硕士学位论文 前1条
1 林岳峰;光学与雷达遥感数据联合反演森林生物量方法研究[D];电子科技大学;2016年
，本文编号：2190670
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2190670.html
