Use of Multi-Feature Extraction and Transfer Learning to Identify Urban Villages in China

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(3), P. 424 - 424

Published: Jan. 26, 2025

Urban villages (UVs) are the most typical urban informal settlements in China, and study of an effective identification method for UVs can help to provide a reference development locally adapted UV transformation policies. In order reduce cost labeling enhance transferability, this integrates remote sensing social data applies sample migration from labeled area less based on theory transfer learning. There two main results study: (1) This constructed feature system multi-feature extraction using block as unit, experiments Tianhe District achieved overall accuracy 90% kappa coefficient 0.76. (2) Using source domain Jiangan target domain, samples were reused KMM, TCA, CORAL algorithms. The CORAL+RF algorithm showed best performance, where its reached 97.06% 0.89, 91.17% 0.67 case no labeling. To sum up, proposed present provides theoretical references methods different geographical areas.

Language: Английский

---

Multi-spatial urban function modeling: A multi-modal deep network approach for transfer and multi-task learning

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2025, Volume and Issue: 136, P. 104397 - 104397

Published: Feb. 1, 2025

Language: Английский

---

Efficient glacial lake mapping by leveraging deep transfer learning and a new annotated glacial lake dataset

Journal of Hydrology, Journal Year: 2025, Volume and Issue: unknown, P. 133072 - 133072

Published: March 1, 2025

Language: Английский

---

MaskPOI: A POI Representation Learning Method Using Graph Mask Modeling

Electronics, Journal Year: 2025, Volume and Issue: 14(7), P. 1242 - 1242

Published: March 21, 2025

Point of Interest (POI) data play a critical role in enabling location-based services (LBS) by providing intrinsic attributes, including geographic coordinates and semantic categories, alongside spatial context that reflects relationships among POIs. However, the inherent label sparsity POI datasets poses significant challenges for traditional supervised learning approaches. To address this limitation, we propose MaskPOI, novel self-supervised framework combines strengths graph neural networks masked modeling. MaskPOI incorporates two complementary modules: an edge mask-based autoencoder models topology predicting existence uncovering hidden feature reconstructs node features to explore rich attribute characteristics Together, these modules enable jointly capture information essential robust representation learning. Extensive experiments demonstrate MaskPOI’s effectiveness improving performance on downstream tasks such as functional zone classification population density prediction. Ablation studies further validate contributions its components, highlighting powerful versatile

Language: Английский

---

Self-supervised representation learning for geospatial objects: A survey

Information Fusion, Journal Year: 2025, Volume and Issue: unknown, P. 103265 - 103265

Published: May 1, 2025

Language: Английский

---

Multimodal Data-Driven Visual Sensitivity Assessment and Planning Response Strategies for Streetscapes in Historic Districts: A Case Study of Anshandao, Tianjin

Land, Journal Year: 2025, Volume and Issue: 14(5), P. 1036 - 1036

Published: May 9, 2025

The landscape visual sensitivity (LVS) assessment is recognized as a critical tool for identifying areas most sensitive to changes and informing multi-resource optimization allocation strategies. However, conventional large-scale LVS criteria methodologies developed natural landscapes do not satisfy the precision-oriented requirements of streetscape (SVS) in historic districts, nor they facilitate operational linkage between outcomes planning applications. This study proposes an innovative SVS–PAP methodology, which systematic integration SVS public esthetic perception (PAP) evaluation. framework was first improved through enriched multi-modal datasets. Subjective weights were obtained via analytic hierarchy process (AHP), incorporating expert judgments, while objective derived entropy weight method (EWM) based on data information entropy. both approaches enhances methodological rigor scientific validity determination. An analytical matrix subsequently constructed assessments PAP-based scenic beauty estimation (SBE), enabling derivation empirical validation conducted Anshandao Historic District yielded four key findings: (1) integrates subjective–objective evaluation factors incorporates broad participation, demonstrates strong reliability, establishing novel paradigm strategic planning; (2) technical framework—leveraging GIS spatial analysis techniques—improves precision, operability, replicability; (3) management strategies formulated by verified reasonable, demonstrating effective planning-transition capability; (4) Notably, historical cultural influences showed significantly higher weighting coefficients across compared non-historic assessments. Overall, these research results address persistent undervaluation spiritual values value trade-off decision-making processes, theoretical practical significance advancement.

Language: Английский

---

Mixed land use measurement and mapping with street view images and spatial context-aware prompts via zero-shot multimodal learning

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2023, Volume and Issue: 125, P. 103591 - 103591

Published: Dec. 1, 2023

Traditional overhead imagery techniques for urban land use detection and mapping often lack the precision needed accurate, fine-grained analysis, particularly in complex environments with multi-functional, multi-story buildings. To bridge gap, this study introduces a novel approach, utilizing ground-level street view images geo-located at point level, to provide more concrete, subtle, informative visual characteristics mixed addressing two major limitations of imagery: coarse resolution insufficient information. Given that spatial context-aware land-use descriptions are commonly employed describe environments, treats as Natural Language Visual Reasoning (NLVR) task, i.e., classifying use(s) based on similarity their local descriptive contexts, by integrating (vision) (language) through vision-language multimodal learning. The results indicate our approach significantly outperforms traditional vision-based methods can accurately capture multiple functionalities ground features. It benefits from incorporation prompts, whereas geographic scale geo-locations matters. Additionally, marks significant advancement mapping, achieving point-level precision. allows representation diverse types locations, offering flexibility various resolutions, including census tracts zoning districts. This is effective areas functionalities, facilitating detailed perspective uses settings.

Language: Английский

---

Multi-modal fusion approaches for tourism: A comprehensive survey of data-sets, fusion techniques, recent architectures, and future directions

Computers & Electrical Engineering, Journal Year: 2024, Volume and Issue: 116, P. 109220 - 109220

Published: April 6, 2024

Language: Английский

---

A Spatial and Adversarial Representation Learning Approach for Land Use Classification with POIs

ACM Transactions on Intelligent Systems and Technology, Journal Year: 2023, Volume and Issue: 14(6), P. 1 - 25

Published: Oct. 17, 2023

Points-of-interests (POIs) have been proven to be indicative for sensing urban land use in numerous studies. However, recent progress mainly relies on spatial co-occurrence patterns among POI categories, which falls short utilizing the rich semantic information embodied hierarchical categories and distribution of POIs at an individual zonal scale. In this context, we present a adversarial representation learning approach (SARL) predicting zones with POIs. SARL deeply mines from both categorical perspectives. Specifically, first utilize convolutional neural network sense each zone. We then leverage autoencoder strategy mine all levels, emphasizes prominent definitive while preserves overall structures Finally, fuse these two perspectives via Wide & Deep carry out prediction fused embeddings. conduct comprehensive experiments validate effectiveness four European cities real-world data. The results demonstrate that substantially outperforms several competitive baselines.

Language: Английский

---

BF-SAM: enhancing SAM through multi-modal fusion for fine-grained building function identification

International Journal of Geographical Information Science, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 27

Published: Sept. 5, 2024

Language: Английский

---