摘要
针对实际应用中高分辨率遥感影像道路提取自动化程度低的现状,提出了一种半自动的高分辨率遥感影像道路提取方法。方法采用数据预处理、尺度分割、分类以及形态优化的工作流程,对高分一号遥感影像进行道路半自动提取。数据预处理利用NDWI、DNVI获得道路潜在区域,边缘增强突出道路边缘信息;采用多尺度分割切割道路潜在区域,尺度对比法获得道路最优分割尺度;主要依据道路的光谱特征、形状特征制定分类规则集进行分类;运用形态学开启运算、闭合运算优化道路形态。实验结果表明:在样本区域内提取精度达到90%,整景影像提取精度达到80%,且可推广到具有陕北地区地貌特征的高分一号影像道路快速提取应用中。
In practical applications, the technology of road extraction from high-resolution remote sensing images is in low automation degree. This paper presents a semi-automatic engineering road extraction method from GF-1 image of Hengshan city, Shanxi province. This method has four main steps including pre-processing, scale segmentation, classification and morphology optimization. Firstly, using NDWI and NDVI to gain the potential road region and edge enhancement to extrude edge information. Secondly, using multiscale segmentation to segment the potential road regions and correlation method to find the optimal segmentation scale. Then,formulating classification rules mostly based on road spectral features and shape features for classification. In the end, morphology opening and closing operation are used to optimize the road extracted. The experimental results show that the extraction precisions are above 90 %from the sample images and above 80 %from the whole image.
出处
《遥感信息》
CSCD
北大核心
2016年第2期64-68,共5页
Remote Sensing Information
基金
高分水利遥感应用示范系统(一期)项目(08-Y30B07-9001-13/15)
关键词
高分一号
预处理
尺度分割
分类规则集
形态优化
. GF-1
pre-processing
scale segmentation
classification rule
morphology optimization
