Computers & Electrical Engineering, Год журнала: 2024, Номер 119, С. 109542 - 109542
Опубликована: Авг. 13, 2024
Язык: Английский
Cities, Год журнала: 2024, Номер 152, С. 105169 - 105169
Опубликована: Июнь 21, 2024
Язык: Английский
Процитировано
38
International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2022, Номер 115, С. 103094 - 103094
Опубликована: Ноя. 12, 2022
Язык: Английский
Процитировано
59
International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2023, Номер 122, С. 103385 - 103385
Опубликована: Июнь 17, 2023
Street View Imagery (SVI) is crucial in estimating indicators such as Sky Factor (SVF) and Green Index (GVI), but (1) approaches terminology differ across fields planning, transportation climate, potentially causing inconsistencies; (2) it unknown whether the regularly used panoramic imagery actually essential for tasks, or we can use only a portion of imagery, simplifying process; (3) do not know if non-panoramic (single-frame) photos typical crowdsourced platforms serve same purposes ones from services Google Baidu Maps their limited perspectives. This study first to examine comprehensively built form metrics, influence different practices on computing them multiple fields, usability normal (from consumer cameras). We overview run experiments 70 million images 5 cities analyse impact multitude variants SVI characterising physical environment mapping street canyons: few (e.g. fisheye) 96 scenarios perspective with variable directions, view, aspect ratios mirroring diverse smartphones dashcams. demonstrate that disparate give mostly comparable results metric R=0.82 R=0.98 metrics); often when using front-facing ultrawide camera), single-frame derive commercial counterparts. finding may simplify processes data also unlock value billions images, which are overlooked, benefit scores locations worldwide yet covered by services. Further, aggregated city-scale analyses, correspond closely.
Язык: Английский
Процитировано
41
Sustainable Cities and Society, Год журнала: 2023, Номер 100, С. 105047 - 105047
Опубликована: Ноя. 8, 2023
Computer vision (CV) technology, a key subset of artificial intelligence, provides powerful tools for extracting valuable insights from visual data, which is crucial component the urban planning process. Despite promising potential CV in planning, its applications this context have not been thoroughly examined. This lack scholarship represents critical knowledge gap our understanding role planning. paper aims to provide consolidated process and challenges planners face during adoption CV. The conducts systematic literature review tackle questions how applied process, what are adopting techniques process? findings revealed: (a) could support broad range tasks including data collection analysis, issue identification prioritisation, public participation, plan design adoption, implementation evaluation; (b) improve decision-making through various information, but limitations need be considered, and; (c) Utilisation efforts sustainable development. study informs policy- plan-making circles by providing into existing prospective contributions transforms augments practices, elaborates adoption.
Язык: Английский
Процитировано
36
Ecological Indicators, Год журнала: 2024, Номер 160, С. 111844 - 111844
Опубликована: Март 1, 2024
Rural landscapes have significant ecological, historical, and cultural value including numerous green spaces forest that should be protected utilized. With the growing demand for tourism in rural areas recent years, greenways become increasingly crucial promoting urban–rural development by connecting linear spatial corridors such as landscape patches, scenic pathways, strip woodlands. The accuracy universality of visual quality assessment affect construction enhancement landscapes. previous studies predominantly focused on individual perceptions greenway landscapes, neglecting comprehensive characteristics both internal composition visible surroundings.The evaluation model usually are purely subjective overall evaluations, lacking exploration prediction mechanisms specific analyses constituent elements. This study correlates percentage elements obtained from image segmentation with evaluation, aims to derive influence ranking space, compares it results objective experiments further improve objectivity conclusions. Based assessment, this utilize deep learning, eye movement analysis, obtain greenways, analyze their influencing factors. A quantitative based components was constructed combining learning information processing human–machine–environment synchronous analysis technology. reveals then relationship between strength attractiveness is consistent human aesthetic preferences, perception, physiological responses. validated scientific, appropriate, precise methodologies quantitatively evaluating preferences
Язык: Английский
Процитировано
16
Land, Год журнала: 2025, Номер 14(2), С. 220 - 220
Опубликована: Янв. 22, 2025
As an important type of linear cultural heritage and a waterfront landscape that integrates both artificial natural elements, canals provide the public with multidimensional perceptual experience encompassing aesthetics, culture, nature. There remains lack refined, micro-level studies on canal landscapes from perspective visual preference. This study focuses typical segment Grand Canal in China, specifically ancient section Yangzhou. We employed SegFormer image semantic segmentation techniques to interpret features 150 panoramic images, quantitatively identifying environmental characteristics canal. Four dimensions were constructed: aesthetic preference, hydrophilic Through questionnaire survey various statistical analyses, we revealed relationships between preferences for characteristics. The main findings include following: (1) Aesthetic preference is positively correlated cultural, natural, preferences, while shows negative correlation preferences. (2) influenced by combination blue-green elements factors. Natural primarily affected increased vegetation visibility, associated higher proportion facilities high-quality pavements, linked larger water surface areas, fewer barriers, better quality. (3) are spatial differences across different urban old city exhibiting aesthetic, than new suburban areas. Finally, this proposes strategies optimising enhancing quality canals, aiming sustainable practical guidance future planning management these sites.
Язык: Английский
Процитировано
2
Ecological Indicators, Год журнала: 2022, Номер 145, С. 109615 - 109615
Опубликована: Ноя. 5, 2022
Gathering knowledge about physical settings and visual information of places has long been interest to a wide variety fields as they affect the experience observers. Previous studies have relied on on-site surveys, low-throughput methods, limited data sources, which especially hinder analyzing waterscape features. Thus, detecting relationships between human perception results large-scale urban water areas waterfront features at high spatial resolutions remains challenging, worldwide not conducted. We investigate an alternative: data-driven waterscapes evaluation approach based computer vision (CV) analyze view imagery (WVI) in 16 cities around world measure how people perceive scenes using virtual reality (VR). bring attention WVI – counterpart street (SVI) bodies, is readily available for many thanks usual SVI services, but entirely overlooked research hitherto. Specifically, deep learning model, trained with 500 segmented water-level photos, was developed them, achieving mean pixel accuracy (MPA) 94%, advances state art. These panoramic images assessed through survey 60 participants indicated their perceptions across multiple dimensions. Afterwards, series statistical analyses were conducted determine indicators that drive perceptions, relationship people's subjective objective environment seen by machines established. The take researchers watercourse planners one step toward understanding interactions semantics globally. dataset we produced this released openly first such instance open imagery, it intended support future studies.
Язык: Английский
Процитировано
34
Sensors, Год журнала: 2023, Номер 23(3), С. 1258 - 1258
Опубликована: Янв. 21, 2023
An appropriate detection network is required to extract building information in remote sensing images and relieve the issue of poor effects resulting from deficiency detailed features. Firstly, we embed a transposed convolution sampling module fusing multiple normalization activation layers decoder based on SegFormer network. This step alleviates missing feature semantics by adding holes fillings, cascading normalizations hold back over-fitting regularization expression guarantee steady parameter classification. Secondly, atrous spatial pyramid pooling decoding fused explore multi-scale contextual overcome issues such as loss local buildings lack long-distance information. Ablation experiments comparison are performed image AISD, MBD, WHU dataset. The robustness validity improved mechanism demonstrated control groups ablation experiments. In comparative with HRnet, PSPNet, U-Net, DeepLabv3+ networks, original algorithm, mIoU dataset enhanced 17.68%, 30.44%, 15.26%, respectively. results show that method this paper superior methods U-Net. Furthermore, it better for integrity edges reduces number false detections.
Язык: Английский
Процитировано
22
Cities, Год журнала: 2024, Номер 156, С. 105473 - 105473
Опубликована: Окт. 21, 2024
Язык: Английский
Процитировано
7
ISPRS Journal of Photogrammetry and Remote Sensing, Год журнала: 2022, Номер 195, С. 90 - 104
Опубликована: Ноя. 29, 2022
Язык: Английский
Процитировано
26