基于GF-1遥感数据的城市绿地信息提取研究  

Information Extraction of urban green land using GF-1 remote sensing data

作　　者：周智勇[1] 邢英梅[1] 东启亮[1] 张志科[1] 胡佳[2] 

机构地区：[1]河北省遥感中心,石家庄050021 [2]中南林业科技大学林业遥感信息工程研究中心,长沙410004

出　　处：《矿产勘查》2015年第5期635-641,共7页Mineral Exploration

摘　　要：城市绿地对城市生态系统具有重要作用,且受到高度重视。为实现城市生态绿地建设与规划,需要进行快速、有效的城市绿地信息提取。随着我国高空间分辨率遥感技术空前发展,使得城市绿地信息提取向着高精度、高效的方向发展。研究拟探索出适用于GF-1遥感数据城市绿地信息提取算法,为城市生态建设提供方法支持。通过对比3种分割方法,找到GF-1数据适用于城市绿地信息提取最佳分割方法与尺度;基于特征分析,对城市绿地进行面向对象规则分类并评价。结果表明,整合NDVI边缘信息为权重的均值漂移分割算法最佳;面向对象规则分类总体精度达到89.00%,kappa系数达到0.8525;相比其他城市绿地类型,防护绿地提取精度仍然存在不足。Urban green plays an important role in the urban ecosystem, and has been highly valued. For the realization of estab- lishing and planning urban ecological green land, quick and effective urban green land information should be extracted. As rapid development of high spatial resolution remote sensing technology in China, urban green land information extraction is developed toward high precision and high efficiency. To find the suitable algorithms for information extraction of urban green land using GF- 1 is to provide ap- proach support for urban ecological construction. After comparing three kinds of segmentation methods, the optimal method and the best segmentation scale for urban green space information extraction by GF-1 data has been found. By the means of feature analysis, the object-oriented classification rules have been classified and evaluated. The results show that the mean shift segmentation algorithm which integrates NDVI edge information as weight is optimal, the overall accuracy of object-oriented classification rules reached 89.00%, and the coefficient of kappa was 0.8525. Compared with other types of urban green space, the extraction accuracy of protection green land remains inadequate.

关 键 词：遥感 GF-1数据 信息提取 城市绿地 

分 类 号：P627[天文地球—地质矿产勘探] S771.8[天文地球—地质学]