Identifying urban villages: an attention-based deep learning approach that integrates remote sensing and street-level images DOI
Sheng Hu,

Zhonglin Yang,

Hanfa Xing

и другие.

International Journal of Geographical Information Science, Год журнала: 2024, Номер unknown, С. 1 - 23

Опубликована: Дек. 17, 2024

Urbanization has been a driving force for economic growth, yet it also caused the emergence of informal urban settlements such as villages (UVs), which are characterized by issues arbitrary land use, high-density construction, and insufficient infrastructure. In previous studies on UV detection, semantic imbalance feature interaction among cross-modal data have not comprehensively considered, impacting accuracy interpretability results. this work, fusion framework is proposed that integrates high-resolution remote sensing street view images detection. First, convolutional neural networks (ResNet-50) used extraction from both images. Then, an inner product channel attention module to dynamically adjust weights while considering multiangle views A incorporates dilation convolution global-based block enhance fusion. The method overall (OA) 0.975 classification in case study Guangzhou–Foshan metropolitan area China, outperforming set baseline methods. integration improves OA value approximately 2%. This work enhances understanding distribution UVs via top-down ground-level automatic efficient way, providing planners with valuable insights accurately identify support targeted, sustainable renewal aligned SDGs inclusive, resilient cities.

Язык: Английский

Deep learning for urban land use category classification: A review and experimental assessment DOI Creative Commons
Ziming Li, Бин Чэн, Shengbiao Wu

и другие.

Remote Sensing of Environment, Год журнала: 2024, Номер 311, С. 114290 - 114290

Опубликована: Июль 14, 2024

Mapping the distribution, pattern, and composition of urban land use categories plays a valuable role in understanding environmental dynamics facilitating sustainable development. Decades effort mapping have accumulated series approaches products. New trends characterized by open big data advanced artificial intelligence, especially deep learning, offer unprecedented opportunities for patterns from regional to global scales. Combined with large amounts geospatial data, learning has potential promote higher levels scale, accuracy, efficiency, automation. Here, we comprehensively review advances based research practices aspects sources, classification units, approaches. More specifically, delving into different settings on learning-based mapping, design eight experiments Shenzhen, China investigate their impacts performance terms sample, model. For each investigated setting, provide quantitative evaluations discussed inform more convincing comparisons. Based historical retrospection experimental evaluation, identify prevailing limitations challenges suggest prospective directions that could further facilitate exploitation techniques using remote sensing other spatial across various

Язык: Английский

Процитировано

31
A graph-based multimodal data fusion framework for identifying urban functional zone DOI Creative Commons
Tao Yuan,

Wanzeng Liu,

Jun Chen

и другие.

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2025, Номер 136, С. 104353 - 104353

Опубликована: Янв. 5, 2025

Язык: Английский

Процитировано

2
An intelligent framework for spatiotemporal simulation of flooding considering urban underlying surface characteristics DOI Creative Commons

Hengxu Jin,

Yiyin Liang,

Haipeng Lu

и другие.

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2024, Номер 130, С. 103908 - 103908

Опубликована: Май 23, 2024

In current urban flood modeling, challenges arise from the inadequate consideration of heterogeneous underlying surface characteristics and complexity parameter optimization processes. This study integrates multiple machine learning methods to propose an intelligent framework for modeling that accounts characteristics. It began by coupling a runoff model with pipe network form interpretable flooding (FM). Subsequently, utilizing BIC-GMM method on sample set model's parameters, this explored grouping trends these parameters. The Transformer was employed classify different categories land use, which, along other environmental indices, aided in construction Artificial Neural Network (ANN) model. expedites acquisition sensitivity also proposes functional zoning rules incorporating "socio-driven-nature-assisted" surface. Finally, clustering feature thresholds sensitive parameters were distributed across various catchment units based area distribution rules. used select observed rainfall-runoff events determine optimal inundation model, culminating BIC-GMM-Transformer-ANN-flooding (BGTA-FM). experimental results indicated reached mean Nash-Sutcliffe efficiency coefficient (NSE) 0.8. performance represents 0.3 0.15 increase NSE compared Transformer-ANN-flooding BIC-GMM-flooding respectively, significantly enhances efficiency. effectively reflects complex environments research area. Our work demonstrates substantial potential integrating physical knowledge reaffirms critical role applying geospatial artificial intelligence (GeoAI) geo-environmental disaster management.

Язык: Английский

Процитировано

3
Identification and Analysis of Ecological Corridors in the Central Urban Area of Xuchang Based on Multi-Source Geospatial Data DOI Creative Commons
Wenyu Wei, Shao‐Hua Wang, Xiao Li

и другие.

ISPRS International Journal of Geo-Information, Год журнала: 2024, Номер 13(9), С. 322 - 322

Опубликована: Сен. 6, 2024

With the development of ecological civilization construction, urban planning and in China have entered a phase which optimizing constructing spaces is required. As national livable city, Xuchang has experienced rapid economic recent years, leading to significant expansion that impacted layout space networks central area its surroundings. Therefore, identifying spatial corridors City are crucial for park city construction. This study utilizes multisource geospatial data identify extract City. Ecological resistance gravity models employed verify primary corridor pattern situated Weidu District, area. Finally, 11 main delineated. In response identification corridors, this integrates analysis methods text evaluate characteristics corridors. The results indicate Xudu Park extends outward, serving as hub network, West Lake Luming form core system. based on relationships, benefits, citizen experience each green parks it traverses, strategies proposed.

Язык: Английский

Процитировано

3
Multi-spatial urban function modeling: A multi-modal deep network approach for transfer and multi-task learning DOI Creative Commons
Zhaoya Gong, Chenglong Wang, Bin Liu

и другие.

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2025, Номер 136, С. 104397 - 104397

Опубликована: Фев. 1, 2025

Язык: Английский

Процитировано

0
SFANet: A Ground Object Spectral Feature Awareness Network for Multimodal Remote Sensing Image Semantic Segmentation DOI Creative Commons
Yizhou Lan, Daoyuan Zheng,

Zheng Ying-jun

и другие.

Remote Sensing, Год журнала: 2025, Номер 17(10), С. 1797 - 1797

Опубликована: Май 21, 2025

The semantic segmentation of remote sensing images is vital for accurate surface monitoring and environmental assessment. Multimodal (RSIs) provide a more comprehensive dimension information, enabling faster scientific decision-making. However, existing methods primarily focus on modality spectral channels when utilizing features, with limited consideration their association to ground object types. This association, commonly referred as the characteristics objects (SCGO), results in distinct responses across different modalities holds significant potential improving accuracy multimodal RSIs. Meanwhile, inclusion redundant features fusion process can also interfere model performance. To address these problems, feature awareness network (SFANet) specifically designed RSIs that effectively leverages by incorporating SCGO proposed. SFANet includes two innovative modules: (1) Spectral Aware Feature Fusion module, which integrates encoder based SCGO, (2) Adaptive Enhancement reduces confusion from information decoder. significantly improves mIoU 5.66% 4.76% compared baseline datasets, outperforming networks adaptively enhanced awareness. demonstrates advancements over other provides new perspectives RSI-specific design characteristics. work offers

Язык: Английский

Процитировано

0
Integrating metro passenger flow data to improve the classification of urban functional regions using a heterogeneous graph neural network DOI Creative Commons
Pengxin Zhang, Min Yang, Yong Wang

и другие.

International Journal of Digital Earth, Год журнала: 2024, Номер 17(1)

Опубликована: Дек. 23, 2024

Язык: Английский

Процитировано

1
LPDi GAN: A License Plate De-Identification Method to Preserve Strong Data Utility DOI Creative Commons
Xiying Li, Liu Heng,

Qunxiong Lin

и другие.

Sensors, Год журнала: 2024, Номер 24(15), С. 4922 - 4922

Опубликована: Июль 30, 2024

License plate (LP) information is an important part of personal privacy, which protected by law. However, in some publicly available transportation datasets, the LP areas images have not been processed. Other datasets applied simple de-identification operations such as blurring and masking. Such crude will lead to a reduction data utility. In this paper, we propose method based on generative adversarial network (LPDi GAN) transform original image synthetic one with generated LP. To maintain attributes, background features are extracted from generate LPs that similar originals. The template style also fed into obtain controllable characters higher quality. results show LPDi GAN can perceive changes environmental conditions tilt angles, control through templates. perceptual similarity metric, Learned Perceptual Image Patch Similarity (LPIPS), reaches 0.25 while ensuring effect character recognition de-identified images, demonstrating achieve outstanding preserving strong

Язык: Английский

Процитировано

0
Identifying urban villages: an attention-based deep learning approach that integrates remote sensing and street-level images DOI
Sheng Hu,

Zhonglin Yang,

Hanfa Xing

и другие.

International Journal of Geographical Information Science, Год журнала: 2024, Номер unknown, С. 1 - 23

Опубликована: Дек. 17, 2024

Urbanization has been a driving force for economic growth, yet it also caused the emergence of informal urban settlements such as villages (UVs), which are characterized by issues arbitrary land use, high-density construction, and insufficient infrastructure. In previous studies on UV detection, semantic imbalance feature interaction among cross-modal data have not comprehensively considered, impacting accuracy interpretability results. this work, fusion framework is proposed that integrates high-resolution remote sensing street view images detection. First, convolutional neural networks (ResNet-50) used extraction from both images. Then, an inner product channel attention module to dynamically adjust weights while considering multiangle views A incorporates dilation convolution global-based block enhance fusion. The method overall (OA) 0.975 classification in case study Guangzhou–Foshan metropolitan area China, outperforming set baseline methods. integration improves OA value approximately 2%. This work enhances understanding distribution UVs via top-down ground-level automatic efficient way, providing planners with valuable insights accurately identify support targeted, sustainable renewal aligned SDGs inclusive, resilient cities.

Язык: Английский

Процитировано

0