摘要
针对利用遥感影像进行森林类型识别容易出现树种误分和模型复杂的问题,以高分一号卫星影像为数据源,结合遥感判读样地、植被指数、纹理信息以及地形因子等多源数据,构建最小距离分类模型、支持向量机分类模型和随机森林分类模型,对黑龙江凉水自然保护区森林优势树种进行分类。结果表明,基于随机森林模型的分类结果总精度和Kappa系数分别为81.01%和0.76,较支持向量机分类方法有明显提高。该研究为提高我国高分辨率数据的自给率和森林资源的有效管理提供了一定的参考价值。
To improve the accuracy of forest tree type identification and reduce the classification model complexity from remote sensing images,this paper tries to propose a new method integrating GF1satellite images with ground survey,vegetation index,texture and terrain factors and other extracted features,and build minimum distance model,support vector machines model and random forest identification model for Liangshui Nature Reserve of Heilongjiang forest dominant tree species classification.The results show that:random forest model achieves an overall accuracy and Kappa coefficient of81.01%and0.76,respectively,which has improved significantly compared with support vector machines method.The proposed method is able to achieve highly satisfactory forest type identification results,improve the self sufficiency rate in using of GF1remotely sensed data and provide an important technical support for the effective management of forest resources.
作者
吕杰
郝宁燕
李崇贵
史晓亮
李宗泽
LV Jie;HAO Ningyan;LI Chonggui;SHI Xiaoliang;LI Zongze(College of Geomatics,Xi'an University of Science and Technology,Xi'an 710054,China)
出处
《遥感信息》
CSCD
北大核心
2017年第6期109-114,共6页
Remote Sensing Information
基金
国家自然科学基金(51409204
41401496)
陕西省自然科学基础研究计划(2015JQ4105)
江西省数字国土重点实验开放基金(DLLJ201604)
陕西省教育厅科研计划项目资助(16JK1496)
2016年陕西省大学生创新创业训练计划项目(20160704068)
关键词
GF-1影像
随机森林
森林类型识别
支持向量机
纹理特征
GF-1 image
random forest
forest type identification
support vector machines
texture feature
