标题: A SegNet-Based Approach for Road Label Placement Integrating Geometric and Textual Information
作者: Yu, HF (Yu, Huafei); Ai, TH (Ai, Tinghua); Yang, M (Yang, Min); Oucheikh, R (Oucheikh, Rachid); Kong, B (Kong, Bo); Wu, H (Wu, Hao); Zhang, ZY (Zhang, Zhenyu); Harrie, L (Harrie, Lars)
来源出版物: TRANSACTIONS IN GIS 卷: 29 期: 4 文献号: e70080 DOI: 10.1111/tgis.70080 Published Date: 2025 JUN
摘要: As a crucial tool for enhancing the readability and comprehensibility of geoinformation, automated label placement in mapping applications remains a significant challenge, particularly when generalizing the label placement process for high-density maps, despite the availability of tools such as QGIS-PAL and ArcGIS-Maplex label engine. This study focuses on utilizing deep learning (DL) for road labeling tasks and addresses two key questions: Can DL models predict the quantity and shape of road labels? Can they determine the label positions? Our proposed SegNet-based model employed "where" and "what" modules, integrating geometric contextual information with textual data as inputs. We validated the model using London, UK wayfinding map data, demonstrating improved readability and achieving comparable machine learning evaluation metrics to mainstream labeling tools. Notably, our method accurately predicted label placements for roads and ensured consistent label sizes. This study provides valuable insights and recommendations for leveraging DL techniques to alleviate labor-intensive challenges of map labeling.
作者关键词: cartographic label placement; deep learning; geometric information; road labels; textual information
KeyWords Plus: NEURAL-NETWORK; MAP; NAMES
地址: [Yu, Huafei; Ai, Tinghua; Yang, Min; Kong, Bo; Wu, Hao; Zhang, Zhenyu] Wuhan Univ, Sch Resource & Environm Sci, Wuhan, Peoples R China.
[Yu, Huafei; Oucheikh, Rachid; Harrie, Lars] Lund Univ, Dept Phys Geog & Ecosyst Sci, Lund, Sweden.
通讯作者地址: Ai, TH (通讯作者)，Wuhan Univ, Sch Resource & Environm Sci, Wuhan, Peoples R China.
Harrie, L (通讯作者)，Lund Univ, Dept Phys Geog & Ecosyst Sci, Lund, Sweden.
电子邮件地址: [email protected]; [email protected]
影响因子：2.3
