基于GF-2影像山地森林有林地小班区划  

Sub-compartment division of mountainous forests with woodland based on GF-2 images

作　　者：师玉霞[1] 袁民 杨勇强 连玲[1] 高亚利 王振锡[1] SHI Yuxia;YUAN Min;YANG Yongqiang;LIAN Ling;GAO Yali;WANG Zhenxi(Collage of Forestry and Horticulture,Xinjiang Agricultural University,Urumqi 830052,Xinjiang,China)

机构地区：[1]新疆农业大学林学与园艺学院,新疆乌鲁木齐830052

出　　处：《中南林业科技大学学报》2022年第9期82-90,共9页Journal of Central South University of Forestry & Technology

基　　金：新疆维吾尔自治区林业改革发展基金项目“新疆天保工程精准监测技术与评价体系研究”(XITB20181102)。

摘　　要：【目的】选用面向对象的分层分类法对研究区域的天山云杉林进行遥感分类,提出高分辨率遥感数据进行有林地小班区划的规则集,检测主流遥感数据源的提取效果,以期为森林资源监测和山区天然林更新恢复评价等方面提供技术参考,进一步推进国产高分辨率遥感数据在山区林地分类方面的应用。【方法】以天山西部山地云杉林为研究对象,以新疆天山西部国有林管理局巩留分局的恰西国家森林公园为研究区,以高分二号(GF-2)影像为基础数据,对其进行多尺度分割和郁闭度估测模型的研究,并结合野外实地测量以及遥感数据的光谱信息、纹理信息、植被指数因子等特征值,建立地物分类规则集,利用面向对象的分层分类方法,开展基于郁闭度的山地森林有林地小班区划研究。【结果】选用地形因子+植被指数+纹理特征因子构建的郁闭度估测模型,估测精度达到了89.93%,能够有效地反映研究区郁闭度的真实情况,而基于郁闭度估测模型的山地森林有林地小班区划的结果较好,根据森林资源二类实地调查结果显示研究区内有林地面积为4264.62 hm^(2),GF-2影像提取出来的有林地面积为4129.82 hm^(2),相对误差只有5.53%,总体吻合度达到了94.47%。【结论】选用国产高分辨率影像进行有林地的提取和郁闭度等级的划分是可行的,可为研究区域内的小班区划提供一定的数据支撑。【Objective】In this article,the object-oriented hierarchical classification method was used to remotely classify the Tianshan spruce forest in the study area,a rule set for forest land sub-class division with high-resolution remote sensing data was proposed,and the extraction effect of mainstream remote sensing data sources was tested.The study has provided technical references for monitoring and evaluating the natural forest regeneration and restoration in mountainous areas,and the application promotion of domestic highresolution remote sensing data in forest land classification in mountainous areas.【Method】This study took the spruce forests in the western Tianshan Mountains as the research object.The Qiaxi National Forest Park of the Gongliu Branch of the State-owned Forestry Administration Bureau in the western Tianshan Mountains of Xinjiang was taken as the research area.Based on the GF-2 images,the multi-scale segmentation and canopy closure estimation models were conducted.Combining with field measurements,spectral information,texture information,vegetation index factors and other characteristic indexes of remote sensing data,a set of classification rules for ground objects was established.In addition,the object-oriented hierarchical classification method was applied to carry out the study on the division of mountain forests with woodland based on canopy density.【Result】The results showed that the canopy closure estimation model constructed by using the combination of terrain factors,vegetation index and texture feature factors,had an estimation accuracy of 89.93%,which can effectively reflect the real situation of canopy closure in the study area.However,the results of the subcompartment division of mountain forests with woodland based on the canopy closure estimation model were better.According to the results of the second-class field survey of forest resources,the area of forest land in the study area was 4264.62 hm^(2),and the area of forest land extracted from the GF-2 images was 4129.82 hm^

关 键 词：GF-2影像 面向对象 多层分类 小班区划 

分 类 号：S771.8[农业科学—森林工程]