利用多源空间数据的城中村空间层次化识别方法

城中村的精细空间分布是城市规划与城市更新的重要参考。由于城中村具有语义高级和遥感影像特征辨识度不足的特点，使用传统的场景识别方法难以从高密度城市中获得精度良好的城中村精细空间分布。针对城中村的精细识别问题，提出了一种新颖的融合遥感影像和社会感知的层次化识别方法。该方法在特征上融合了遥感图像和社会感知数据的优点，其层次化结构同时考虑了大范围的上下文信息和小范围的局部信息，为在精细尺度全面理解城中村提供了一个新思路。基于该方法对深圳市的城中村进行了空间识别，获得了2.5 m空间分辨率的精细城中村分布。精度验证表明，该结果的总体精度和Kappa系数分别达到98.68%和0.807，说明该方法具有优秀的表现。此外，还通过对照实验分别证明了层次化识别框架、融合遥感影像和社会感知数据的增益效果。结果表明，层次化框架和多源空间数据都能有效提高城中村识别方法的精度。

Hierarchical Spatial Recognition Method for Urban Villages by Integrating Multi-source Geospatial Data

  Objectives  The fine spatial distribution of urban villages is important for urban planning and urban renewal. However, since urban villages are high-level semantic geo-objects and have obscure remote sensing characteristics, it is difficult to obtain fine spatial distribution with good precision from high-density cities using traditional methods.  Methods  We propose a novel hierarchical recognition method for urban villages that fuses remote sensing images and social sensing data to finely recognize the urban villages. The method combines the advantages of remote sensing images and social perception data in features. Large- and small-scale information are both considered into the process by using the hierarchical framework.  Results  The method provides a new idea for a comprehensive understanding of urban villages at a fine scale. A case study has been implemented in Shenzhen. An urban village distribution with a spatial resolution of 2.5 m is obtained. The accuracy assessment shows that the overall accuracy and Kappa coefficient reach 98.68% and 0.807, respectively, indicating the excellent performance of the method. In addition, the gain effects of the hierarchical framework and the fusion of remote sensing images and social perception data are demonstrated, respectively.  Conclusions  The results show that both the hierarchical framework and the multi-source spatial data are effective in improving the accuracy of the urban village recognition method.