摘要
基于无人机可见光影像对古尔班通古特沙漠地表类型信息进行提取,运用面向对象的多尺度分割,在提取样本的光谱、形状、纹理、植被指数特征的基础上,建立规则提取地表类型信息。结果表明:(1)荒漠地表类型不同,最佳分割尺度不同;(2)不同荒漠化程度地表类型特征相似,无法运用单个特征进行区分,需选用多种特征组合提取地表类型;(3)面向对象的多尺度分割方法相对于基于像元的最大似然法分类有明显提高,面向对象轻度沙漠区总体分类精度为93. 00%,中度沙漠化区为91. 83%,重度沙漠化区为93. 50%,较基于像元的最大似然方法分别提高了10. 34%、11. 86%和12. 50%。表明针对无人机可见光影像,面向对象的多尺度分割方法能高精度地提取荒漠地表类型信息。
Based on the images of unmanned aerial vehicle(UAV),in this study the surface type information of the Gurbantunggut Desert was extracted,and the object-oriented multi-scale segmentation was used to extract the information of the sample plots of surface types from the spectrum,shape,texture and vegetation index.The results showed that:(1)The best segmentation scale for the different desert surface types was different;(2)The features of surface types with different desertification levels were similar and could not be distinguished by single characteristics;(3)Compared with the pixel-based maximum likelihood method,the object-oriented multi-scale segmentation method was improved significantly.The overall classification accuracies of the object-oriented slightly,moderately and seriously desertified areas were 93.00%,91.83%and 93.50%respectively,and they were 10.34%,11.86%and 12.50%higher than those of the pixel-based maximum likelihood method.The results revealed that the objectoriented multi-scale segmentation method could be used to extract the desert surface type information with high accuracy for the UAV visible image.
作者
彭佳忆
王新军
朱磊
赵成义
徐晓龙
PENG Jia-yi;WANG Xin-jun;ZHU Lei;ZHAO Chen-yi;XU Xiao-long(College of Grassland and Environmental Sciences,Xinjiang Agricultural University,Urumqi 830052,Xinjiang,China;Xinjiang Key Laboratory of Soil and Plant Ecological Processes,Urumqi 830052,Xinjiang,China;Xinjiang Institute of Ecology and Geography,Chinese Academy of Sciences,Urumqi 830011,Xinjiang,China)
出处
《干旱区研究》
CSCD
北大核心
2019年第3期771-780,共10页
Arid Zone Research
基金
国家自然科学基金项目(41761085
41301205)资助
