摘要
为快速准确提取可见光遥感图像中的林区植被,降低林区复杂地物与不均匀的光照对提取效果的影响。采用无人机获取的林区可见光遥感图像,利用ArcGIS软件根据植被与裸地、道路以及光照均匀程度的不同占比进行裁剪,获得5个试验样区,分别利用多尺度分割、光谱差异分割和多尺度结合光谱差异分割方法对样区影像进行分割,应用最近邻分类方法分类并分析3种分割方法对分类精度的影响。研究结果表明:基于多尺度分割的分类精度整体优于光谱差异分割和多尺度结合光谱差异分割,植被分类总体精度分别为90.0%、93.0%、92.0%、89.0%、94.0%,Kappa系数分别为0.801、0.855、0.839、0.781、0.880。使用多尺度分割在林区植被提取时受环境影响小,可以有效提取林区植被信息。
The experiment was conducted to quickly and accurately extract forest vegetation in visible light remote sensing images,and reduce the influence of complex ground objects and uneven illumination in forest areas on the extraction effect.The visible light remote sensing images of forest areas obtained by drones were used.According to the different proportions of vegetation and bare land,roads and illumination uniformity,ArcGIS software was used to cut out five experimental sample areas.Multi-scale segmentation,spectral difference segmentation and multi-scale segmentation combined spectral difference segmentation methods were used to segment the plot images,and the nearest neighbor classification method was used to classify and analyze the influence of the three segmentation methods on the classification accuracy.The classification accuracy of multi-scale segmentation is generally better than that of spectral difference segmentation and multi-scale segmentation combined spectral difference segmentation.The overall accuracy of vegetation classification is 90.0%,93.0%,92.0%,89.0%and 94.0%,respectively,and the Kappa coefficient is 0.801,0.855,0.839,0.781 and 0.880.Multi-scale segmentation is less affected by the environment in forest vegetation extraction,and it can effectively extract forest vegetation information.
作者
刘旭光
肖啸
兰玉彬
苗建驰
巩道财
赵静
张国富
Liu Xuguang;Xiao Xiao;Lan Yubin;Miao Jianchi;Gong Daocai;Zhao Jing;Zhang Guofu(Shandong University of Technology,Zibo 255000,P.R.China;Forest Resources Monitoring and Protection Service Center,Qixia City)
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2023年第4期62-67,共6页
Journal of Northeast Forestry University
基金
山东省引进顶尖人才“一事一议”专项经费项目(鲁政办字[2018]27号)
山东省自然科学基金项目(ZR2021MD091)。
关键词
无人机遥感
多尺度分割
光谱差异分割
松树植被提取
最近邻分类
UAV remote sensing
Multi-scale segmentation
Spectral difference segmentation
Pine vegetation extraction
Nearest neighbor classification
