摘要
针对高空间分辨率卫星遥感数据多光谱波段数较少和原始波段组合光谱特征利用有限等问题,该文提出了一种基于特征变换的建筑物信息提取方法。以陕西省延安市宝塔地区为研究区,基于快鸟数据采取特征变换、波段选择、数据融合等解决高空间分辨率原始光谱特征利用有限等问题,采用知识规则的面向对象分类方法进行建筑物识别研究。实验表明,缨帽变换波段能有效地突出建筑物信息,4种融合算法中主成分变换融合适用进一步面向对象分类,建筑物识别的总体精度达到89.3%。此方法能有效识别沿坡脚或滑坡体分散分布的建筑物,为快速获取居民空间分布信息和辅助灾害应急评估等提供参考。
According to the problem of the fewer original spectral bands and limited spectral characteristics of high spatial resolution satellite remote sensing data,this paper proposed a method for building extraction based on feature transformation.In this study,we examined how building identification can be realized through object-oriented classification based on knowledge rules by bands selection and feature transformation to the Quickbird multispectral data of Yan’an Baota,Shaanxi province,Northwest China,and then applied data fusion by using four different algorithms.The experimental results suggested that the tasseled cap(TC)transformation effectively highlights the spectral information features of buildings;furthermore,principal component fusion can be applied to object-oriented classification with an overall accuracy of 89.3%.This method effectively identifies buildings irregularly distributed along slope toes or landslide masses,provided reference for quickly acquiring the spatial distribution information of local residents,and supported disaster emergency response assessment.
作者
丁辉
张茂省
DING Hui;ZHANG Maosheng(Xi'an Center of Geological Survey,China Geological Survey,Xi'an 710054,China;Key Laboratory for Geo-hazard in Losses Area,Ministry of Land and Resources,Xi'an 710054,China)
出处
《测绘科学》
CSCD
北大核心
2019年第1期124-131,143,共9页
Science of Surveying and Mapping
基金
国家自然科学基金项目(41502338
41530640)
中国地质调查局地质调查项目(12120114025701)
关键词
缨帽变换
快鸟
建筑物
面向对象
信息识别
Tasseled cap transformation
Quickbird
building
object-oriented
information identification
