摘要
采用长春市WorldView-3高分辨率遥感影像数据,利用面向对象方法建立多尺度网络层,同时基于不同地物的光谱、几何特征构建了eCognition中的分类规则集,提取了城市下垫面用地信息。最后通过混淆矩阵对分类结果进行精度评价,得到较好的分类效果。本次研究初步解决了阴影归类和建筑与硬质地面区分等问题,提高了城市下垫面信息提取精度。
In this paper, based on the WorldView-3 high-resolution remote sensing image data of Changchun City, we used the object-oriented method to establish the multi-scale network layer. At the same time, we constructed the classification rule set in e-Cognition to extract the information of urban underlying surface based on the spectral and geometric features of different objects. Finally, we evaluated the classification result by the confusion matrix. This study can solve the problems of shadow classification and the distinction between buildings and ground preliminarily, and improve the precision of information extraction of urban underlying surface.
出处
《地理空间信息》
2020年第4期100-104,I0008,共6页
Geospatial Information
基金
吉林省科技厅重点科技攻关项目(20170204036SF)。
关键词
高分辨率影像
面向对象
城市下垫面
影像解译
易康
high-resolution image
object-oriented
urban underlying surface
image interpretation
e-Cognition