基于面向对象和多尺度影像分割技术的城市用地分类研究——以武汉市城市中心区为例

Urban Land Cover Classification Research Based on Object-oriented and Multi-scale Image Segmentation Techniques——A Case Study of Downtown Area of Wuhan City

对城市用地进行分类研究,可为城市土地的合理规划布局与有效管理提供即时信息支撑。利用Quik-Bird高分辨率遥感影像,采用面向对象的影像分析技术与多尺度影像分割技术,运用eCogni-tion软件对武汉市城市中心区用地进行分类研究。针对研究区各用地类型,构建优化分类模型,选择适宜的影像分割尺度和分割特征、按照合理的顺序逐步提取城市用地分类信息。评价结果表明:该方法分类速度快、精度高,为高分辨率遥感影像城市用地分类研究提供了新思路。

The classified research into urban land prov ides the rational layout and efficient management with instant information support. Using Quik-Bird high resolution remote sensing images and object-oriented and multi-scale image segmentation and adopting the eCognition software, this paper conducts classified research into the downtown area of Wuhan City. According to the different types of the land in the city, the researchers create the module of classification, select proper scale parameter and segmentation characters, and then take proper sequence to extract the urban land cover information. Accuracy assessment result indicates that this method has a precise result and short course,providing a new way for the research on the urban land cover classification.