Recognizing urban functional zones by a hierarchical fusion method considering landscape features and human activities. Issue 5 (17th June 2020)

• Record Type: Journal Article
• Authors: Liu, Huimin; Xu, Yiyuan; Tang, Jianbo; Deng, Min; Huang, Jincai; Yang, Wentao; Wu, Fang
• Is Part Of: Transactions in GIS. Volume 24:Issue 5(2020)
• Page Start: 1359
• Page End: 1381
• Publication Date: 2020-06-17
• Subjects: Geographic information systems -- Periodicals; 910.285
• DOI: 10.1111/tgis.12642
• Languages: English
• ISSNs: 1361-1682

Abstract

Abstract: Functional zones are basic spatial units of urban management and urban planning and they provide the basic space where human activities occur. The concept of urban functional zones is closely related to two basic factors: urban landscape environment and human activities. The former provides the basic space for human activity and influences urban land use at a coarse scale, and the latter indicates the differentiation of functions in local urban areas. In previous studies, the hierarchical correspondence and interaction between urban landscape and human activities have not been given full consideration in the cognition of urban functional zones, which would influence the accuracy and interpretability of the results. Therefore, a hierarchical fusion method considering urban landscape and human activity patterns based on multi‐source data is proposed in this article. We introduce a new cognition framework. First, a land use classification based on urban landscape features from remote sensing images is extracted and used as the fundamental results of urban land uses. Then, fine‐grained functional semantics of local urban areas are recognized based on human activity patterns extracted from crowdsourced data (i.e., points of interest and taxi trajectories). Finally, the above results at different scales are fused with hierarchical constraints, which means that the information from both urban landscape and human activities can be fully utilized to obtain a refined and fine‐grained map of urban functional zones. Wuhan, China was chosen as an example to evaluate the performance of the proposed method. Results show that the overall accuracy of the proposed method is 82.51% (accuracies of the mixed functional zones and single functional zones are 77.93 and 87.96%, respectively) and compared with state‐of‐the‐art methods, the proposed method performed better for the recognition of mixed functional zones.