基于GF-1的森林蓄积量遥感估测

详细信息    查看全文 | 推荐本文 |

英文篇名：Estimation of forest volume based on GF-1 作者：李世波 ; 林辉 ; 王光明 ; 程韬略 英文作者：LI Shibo;LIN Hui;WANG Guangming;CHENG Taolue;Research Center of Forestry Remote Sensing & Information, Central South University of Forestry and Technology;Guizhou Forestry Reconnaissance and Design Co.Ltd;Liling City Forestry Bureau; 关键词：森林蓄积量 ; 遥感因子 ; 多元逐步回归 ; 偏最小二乘回归 ; 随机森林模型 英文关键词：forest stock volume;;remote sensing factor;;multiple stepwise regression;;partial least-squares;;random forest model 中文刊名：ZNLB 英文刊名：Journal of Central South University of Forestry & Technology 机构：中南林业科技大学林业遥感信息工程研究中心;贵州林业勘察设计有限公司;醴陵市林业局; 出版日期：2019-06-25 16:13 出版单位：中南林业科技大学学报 年：2019 期：v.39;No.218 基金：“十三五”国家重点研发计划子课题“单木-林分尺度人工林资源遥感精细检测技术”(2017YFD0600902);; 湖南省科技厅项目“林业遥感大数据与生态安全”(2016TP1014) 语种：中文; 页：ZNLB201908011 页数：7 CN：08 ISSN：43-1470/S 分类号：75-80+91

摘要

森林蓄积量是评价森林资源数量的一个重要指标。结合遥感影像和地面调查数据估测森林蓄积量受遥感影像、遥感因子、预处理方法、估测方法等多方面的影响。为研究国产GF-1遥感影像估测森林蓄积量的最佳遥感因子组合方式和较优估测方法,并绘制森林蓄积量空间分布图,为我国森林蓄积量的研究提供理论基础和科学依据。为研究GF-1遥感影像估测森林蓄积量的遥感因子和估测方法,以湖南省醴陵市为研究对象,以国产GF-1遥感影像为数据源,通过对遥感图像预处理,获取光谱信息、纹理因子、植被指数作为特征变量,结合同时期的二类调查样地数据,从GF-1遥感影像像元与样地不匹配角度出发,应用移动窗口的方法解决像元与样地的对应关系,采用多元逐步回归、偏最小二乘回归和随机森林模型对研究区森林蓄积量进行估测,采用建模精度和估测精度进行分析评价。实验结果表明:1)3个模型选择的因子都包含了NDVI、 Band2、DI3、CO1和DVI等5个遥感因子,说明其对森林蓄积量的估测比较敏感;2)随机森林模型优于偏最小二乘回归和多元逐步回归,其决定系数R2为0.73、估测精度为83.69%。利用GF-1遥感影像结合随机森林模型应用于森林蓄积量的估测结果趋于真实分布,效果较理想;采用移动窗口法,利用国产GF-1遥感影像并结合随机森林进行森林蓄积量估测具有较好的应用前景。
        Forest stock volume is an important index to evaluate the quantity of forest resources. Forestry volume estimation based on remote sensing image and ground survey data is affected by remote sensing image, remote sensing factors, preprocessing methods, estimation methods and so on. In order to study the best combination of remote sensing factors and the best estimation method for estimating forest stock with domestic GF-1 remote sensing image, and to draw the spatial distribution map of forest stock, provide theoretical basis and scientific basis for the study of forest stock in China. In this study, we studied the remote sensing factors and methods for estimating forest stock using GF-1 remote sensing images. Taking Liling City, Hunan Province as the research object, and using domestic GF-1 remote sensing images as data sources, we obtained spectral information, texture factor and vegetation index as characteristic variables through remote sensing image preprocessing. Combining with two types of survey sample data in the same period, we obtained the remote sensing image pixels and sample plots from GF-1 remote sensing image. From the mismatch point of view, the method of moving window is applied to solve the corresponding relationship between pixels and sample plots. The forest volume in the study area is estimated by multiple stepwise regression, partial least squares regression and random forest model, and the accuracy of modeling and estimation is analyzed and evaluated. The results showed that:1) the selected factors of the three models include five remote sensing factors, NDVI, Band2, DI3, CO1 and DVI, which indicate that they are sensitive to the estimation of forest stock; 2) random forest model is superior to partial least squares regression and multiple stepwise regression. Its determinant coefficient R~2 is 0.73 and its estimation accuracy is 83.69%. The results show that the application of GF-1 remote sensing image combined with Stochastic Forest Model in forest stock estimation tends to be true distribution, and the effect is ideal. The application of moving window method and domestic GF-1 remote sensing image combined with Stochastic Forest in forest stock estimation has good application prospects.

引文

[1]施赵君.基于遥感信息的森林蓄积量估测[D].杭州:浙江农林大学,2014.
    [2]王佳,宋珊芸,刘霞,等.结合影像光谱与地形因子的森林蓄积量估测模型[J].农业机械学报,2014,45(5):216-220.
    [3]刘琼阁,彭道黎,涂云燕.基于偏最小二乘回归的森林蓄积量遥感估测[J].中南林业科技大学学报,2014,34(2):81-84,132.
    [4]BEAUDOIN A,LE TOAN T,GOZE S,et al.Retrieval of forest biomass from SAR data[J].International Journal of Remote Sensing,1998(15):2777-2796.
    [5]HOUGHTON R A,BUTMAN D,BUNN A G,et al.Mapping Russian forest biomass with data from satellites and forest inventories[J].Environmental Research Letters,2007,2(4):45032-45037.
    [6]FAZAKAS Z,NILSSON M,OLSSON H.Regional forest biomass and wood volume estimation using satellite data and ancillary data[J].Agricultural and Forest Meteorology,1999(98):417-425.
    [7]张超,彭道黎,涂云燕,等.利用TM影像和偏最小二乘回归方法估测三峡库区森林蓄积量[J].北京林业大学学报,2013,35(3):11-17.
    [8]洪奕丰,林辉,严恩萍,等.基于偏最小二乘法的平南县森林蓄积量估测模型研究[J].中南林业科技大学学报,2011,31(7):80-85.
    [9]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.
    [10]刘鑫.基于RS和GIS的森林蓄积量估测的研究[D].南京:南京林业大学,2010.
    [11]张彦林.基于3S技术的山东省森林蓄积量定量估测研究[D].北京:北京林业大学,2008.
    [12]琚存勇,蔡体久.用泛化改进的BP神经网络估测森林蓄积量[J].林业科学,2006,42(12):59-62.
    [13]涂云燕,张盼盼.基于SPOT-5的森林蓄积量估测模型研究[J].林业建设,2015(4):61-65.
    [14]陈尔学,李增元,武红敢,等.基于k-NN和Landsat数据的小面积统计单元森林蓄积估测方法[J].林业科学研究,2008,21(6):745-750.
    [15]ST-ONGE B,HU Y,VEGA C.Mapping the height and aboveground biomass of a mixed forest using lidar and stereo Ikonos images[J].International Journal of Remote Sensing,2008,29(5):1277-1294.
    [16]L I D D,FENG Z K,WANG X A,et al.Model of BP neural network inversing forest biomass[J].Forest Inventory&Planning,2013,38(1):5-8.
    [17]郑刚.基于KNN法的森林蓄积量的遥感估计和反演[D].南京:南京林业大学,2009.
    [18]郭含茹,张茂震,徐丽华,等.基于地理加权回归的区域森林碳储量估计[J].浙江农林大学学报,2015,32(4):497-508.
    [19]戚玉娇,李凤日.基于KNN方法的大兴安岭地区森林地上碳储量遥感估算[J].林业科学,2015,51(5):46-55.
    [20]王月婷,张晓丽,杨慧乔,等.基于Landsat8卫星光谱与纹理信息的森林蓄积量估算[J].浙江农林大学学报,2015,32(3):384-391.