A Large-Scale Multipurpose Benchmark Dataset and Real-Time Interpretation Platform Based on Chinese Rural Buildings DOI Creative Commons
Weihuan Deng, Weipan Xu, Yaofu Huang

и другие.

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Год журнала: 2024, Номер 17, С. 10914 - 10928

Опубликована: Янв. 1, 2024

As urbanization accelerates, the evolving dynamics of village growth and decline have garnered widespread attention. Rural housing, as most significant asset in villages, serves primary indicator socio-economic development rural areas. However, extensive scale, diversity, distribution villages make conducting a nationwide census buildings notably costly time-intensive endeavor. Although deep learning techniques been successfully applied by numerous researchers to map building footprints, majority this work is concentrated urban areas, leaving large-scale datasets for lacking. In article, an exhaustive database architecture has established, featuring diverse annotations from provinces mainland China. Moreover, real- time online platform remote sensing image interpretation, integrating instance segmentation boundary regularization, developed streamline extraction footprints high-resolution imagery. Experimental results predicting 43,992 instances demonstrated that 33,210 were accurately identified, achieving precision 0.776, recall 0.755, F1 score 0.765. Building upon work, maps areas quantity are produced clearly demonstrate houses parts These data products can serve vital supplements public such nighttime light data, land cover maps, national statistical yearbooks, road network particularly field studies.

Язык: Английский

Deep learning for cross-domain data fusion in urban computing: Taxonomy, advances, and outlook DOI
Xingchen Zou, Yibo Yan, Xixuan Hao

и другие.

Information Fusion, Год журнала: 2024, Номер 113, С. 102606 - 102606

Опубликована: Авг. 5, 2024

Язык: Английский

Процитировано

18
Urbanization enhances channel and surface runoff: A quantitative analysis using both physical and empirical models over the Yangtze River basin DOI
Shuzhe Huang,

Yuan Gan,

Nengcheng Chen

и другие.

Journal of Hydrology, Год журнала: 2024, Номер 635, С. 131194 - 131194

Опубликована: Апрель 6, 2024

Язык: Английский

Процитировано

17
Learning visual features from figure-ground maps for urban morphology discovery DOI
Jing Wang, Weiming Huang, Filip Biljecki

и другие.

Computers Environment and Urban Systems, Год журнала: 2024, Номер 109, С. 102076 - 102076

Опубликована: Фев. 3, 2024

Язык: Английский

Процитировано

15
A multimodal data fusion model for accurate and interpretable urban land use mapping with uncertainty analysis DOI Creative Commons
Xiaoqin Yan, Zhangwei Jiang,

Peng Luo

и другие.

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2024, Номер 129, С. 103805 - 103805

Опубликована: Апрель 4, 2024

Urban land use patterns can be more accurately mapped by fusing multimodal data. However, many studies only consider socioeconomic and physical attributes within parcels, neglecting spatial interaction uncertainty caused To address these issues, we constructed a data fusion model (MDFNet) to extract natural physical, socioeconomic, connectivity ancillary information from We also established an analysis framework based on generalized additive learnable weight module explain data-driven uncertainty. Shenzhen was chosen as the demonstration area. The results demonstrated effectiveness of proposed method, with test accuracy 0.882 Kappa 0.858. Uncertainty indicated contributions in overall task 0.361, 0.308, 0.232 for remote sensing, social taxi trajectory data, respectively. study illuminates collaborative mechanism various categories, offering accurate interpretable method mapping urban distribution patterns.

Язык: Английский

Процитировано

12
UrbanCLIP: Learning Text-enhanced Urban Region Profiling with Contrastive Language-Image Pretraining from the Web DOI
Y. H. Yan, Haomin Wen, Siru Zhong

и другие.

Proceedings of the ACM Web Conference 2022, Год журнала: 2024, Номер unknown, С. 4006 - 4017

Опубликована: Май 8, 2024

Язык: Английский

Процитировано

9
Zero-shot urban function inference with street view images through prompting a pretrained vision-language model DOI
Weiming Huang, Jing Wang, Gao Cong

и другие.

International Journal of Geographical Information Science, Год журнала: 2024, Номер 38(7), С. 1414 - 1442

Опубликована: Май 22, 2024

Inferring urban functions using street view images (SVIs) has gained tremendous momentum. The recent prosperity of large-scale vision-language pretrained models sheds light on addressing some long-standing challenges in this regard, for example, heavy reliance labeled samples and computing resources. In paper, we present a novel prompting framework enabling the model CLIP to effectively infer fine-grained with SVIs zero-shot manner, that is, without training. UrbanCLIP comprises an taxonomy several function prompt templates, order (1) bridge abstract categories concrete object types can be readily understood by CLIP, (2) mitigate interference SVIs, street-side trees vehicles. We conduct extensive experiments verify effectiveness UrbanCLIP. results indicate largely surpasses competitive supervised baselines, e.g. fine-tuned ResNet, its advantages become more prominent cross-city transfer tests. addition, UrbanCLIP's performance is considerably better than vanilla CLIP. Overall, simple yet effective inference, showcases potential foundation geospatial applications.

Язык: Английский

Процитировано

8
A review of crowdsourced geographic information for land-use and land-cover mapping: current progress and challenges DOI Creative Commons
Hao Wu, Yan Li, Anqi Lin

и другие.

International Journal of Geographical Information Science, Год журнала: 2024, Номер 38(11), С. 2183 - 2215

Опубликована: Июль 14, 2024

The emergence of crowdsourced geographic information (CGI) has markedly accelerated the evolution land-use and land-cover (LULC) mapping. This approach taps into collective power public to share spatial information, providing a relevant data source for producing LULC maps. Through analysis 262 papers published from 2012 2023, this work provides comprehensive overview field, including prominent researchers, key areas study, major CGI sources, mapping methods, scope research. Additionally, it evaluates pros cons various sources methods. findings reveal that while applying with labels is common way by using analysis, limited incomplete coverage other quality issues. In contrast, extracting semantic features interpretation often requires integrating multiple datasets remote sensing imagery, alongside advanced methods such as ensemble deep learning. paper also delves challenges posed in explores promising potential introducing large language models overcome these hurdles.

Язык: Английский

Процитировано

7
High-resolution mapping of GDP using multi-scale feature fusion by integrating remote sensing and POI data DOI Creative Commons

Nan Wu,

Jining Yan, Dong Liang

и другие.

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2024, Номер 129, С. 103812 - 103812

Опубликована: Апрель 9, 2024

High-resolution spatial distribution maps of GDP are essential for accurately analyzing economic development, industrial layout, and urbanization processes. However, the currently accessible gridded datasets limited in number resolution. Furthermore, high-resolution mapping remains a challenge due to complex sectoral structure GDP, which encompasses agriculture, industry, services. Meanwhile, multi-source data with high resolution can effectively reflect level regional development. Therefore, we propose multi-scale fusion residual network (Res-FuseNet) designed estimate grid density by integrating remote sensing POI data. Specifically, Res-FuseNet extracts features relevant different sectors. It constructs joint representation through mechanism estimates three sectors using connections. Subsequently, obtained correcting overlaying each sector county-level statistical The 100-meter map urban agglomeration middle reaches Yangtze River 2020 was successfully generated this method. experimental results confirm that outperforms machine learning models baseline model significantly training across at town-level. R2 values 0.69, 0.91, 0.99, respectively, while town-level evaluation also exhibit accuracy (R2=0.75). provides an innovative method, reveal characteristics structures fine-scale disparities within cities, offering robust support sustainable

Язык: Английский

Процитировано

6
Context-aware multi-head self-attentional neural network model for next location prediction DOI Creative Commons
Ye Hong, Yatao Zhang, Konrad Schindler

и другие.

Transportation Research Part C Emerging Technologies, Год журнала: 2023, Номер 156, С. 104315 - 104315

Опубликована: Сен. 11, 2023

Accurate activity location prediction is a crucial component of many mobility applications and particularly required to develop personalized, sustainable transportation systems. Despite the widespread adoption deep learning models, next models lack comprehensive discussion integration mobility-related spatio-temporal contexts. Here, we utilize multi-head self-attentional (MHSA) neural network that learns transition patterns from historical visits, their visit time duration, as well surrounding land use functions, infer an individual's location. Specifically, adopt point-of-interest data latent Dirichlet allocation for representing locations' contexts at multiple spatial scales, generate embedding vectors features, learn predict with MHSA network. Through experiments on two large-scale GNSS tracking datasets, demonstrate proposed model outperforms other state-of-the-art reveal contribution various model's performance. Moreover, find trained population achieves higher performance fewer parameters than individual-level due collective movement patterns. We also conducted in recent past one week before has largest influence current prediction, showing subset sufficient obtain accurate result. believe vital context-aware prediction. The gained insights will help understand promote implementation applications.

Язык: Английский

Процитировано

12
Mixed land use measurement and mapping with street view images and spatial context-aware prompts via zero-shot multimodal learning DOI Creative Commons
Meiliu Wu, Qunying Huang, Song Gao

и другие.

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2023, Номер 125, С. 103591 - 103591

Опубликована: Дек. 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.

Язык: Английский

Процитировано

10