Building extraction is one of the important research directions that has attracted great attention in the field of remote sensing image processing. It refers to the process of accurately extracting building information such as the location and shape of buildings by analyzing and processing remote sensing images. This technology plays an irreplaceable and important role in urban planning, disaster management, map production, smart city construction, and other fields. In recent years, with the advancement of science and technology, especially the continuous evolution of earth observation technology and the rapid development of deep learning algorithms, Convolutional Neural Networks (CNNs) have become an emerging solution for extracting buildings from remote sensing images because of their powerful feature extraction capability. The aim of this paper is to provide a comprehensive and systematic overview and analysis of building extraction methods based on convolutional neural networks. We conduct a comprehensive literature review to summarize the building extraction methods from perspectives of model structure, multi-scale feature differences, lack of boundary information, and model complexity. This will help researchers to better understand the advantages and disadvantages of different methods and the applicable scenarios. In addition, several typical building datasets in this field are described in detail, as well as the potential issues associated with these datasets. Subsequently, by collecting experimental results of relevant algorithms on these typical datasets, a detailed discussion on the accuracy and parameter quantities of various methods is conducted, aiming to provide a comprehensive assessment of performance and applicability of these methods. Finally, based on the current research status of this field and looking forward to the new era of high-quality development in artificial intelligence, the future directions for building extraction are prospected. Specifically, this paper discusses the combination of Transformers and CNNs, the combination of deep learning and reinforcement learning, multi-modal data fusion, unsupervised or semi-supervised learning methods, real-time extraction based on large-scale remote sensing model, building instance segmentation, and building contour vector extraction. In conclusion, our review can provide some valuable references and inspirations for future related research, so as to promote the practical application and innovation of building extraction from remote sensing images. This will fulfill the demand for efficient and precise map information in remote sensing technology and other related fields, contributing to the sustainable and high-quality development of human society.

With the expansion of urban areas, a mix of transportation modes has become prevalent during the daily commutes of city dwellers. That is, commuters often need to transfer between various modes to reach their destinations. Accurate identification and analysis of these transfer behaviors are crucial for advancing urban transportation research. Current research tends to focus on distance or time thresholds, typically derived from walking speeds or anecdotal experience. However, these approaches often overlook the distinct station densities within cities. Other studies, while utilizing GPS, GTFS, and similar datasets, construct intricate transfer identification methods that lack generalizability. Against this backdrop, we introduce a time-distance dual-constraint transfer recognition algorithm. Firstly, leveraging extensive traffic IC card data, based on the statistical characteristics of the proximity distance sequences between bus or subway stations and their M neighboring stations, distance thresholds for bus-bus, bus-subway, and subway-bus transfer are detected individually. Subsequently, a filtering algorithm based on these distance thresholds is applied to daily data to produce a candidate transfer data set. Based on this, four time thresholds for each day are determined by analyzing the statistical characteristics of the transit time differences within the datasets. Finally, these dual thresholds facilitate the precise extraction of transfer behaviors. Furthermore, we establish a classification framework for these behaviors, classifying them into nine distinct transfer modes. These modes are defined based on the duration of travel time in the first and second journeys, encompassing variations including long-long, long-medium, long-short, middle-long, middle-middle, middle-short, short-long, short-middle, and short-short. We analyze these models individually for their travel characteristics. Results reveal that the morning peak for all transfer trips precedes that of buses and subways, with short-long transfers leading by up to 30 minutes. This underscores the added effort required by commuters who rely on transfers. In contrast, evening peak times vary, with certain transfer modes like long-long and long-short lagging notably behind the general evening peak. This further emphasizes the increased commuting burden associated with transfers. In terms of travel distances, the peak of regular subway travel distances is around 10 km, while that of the bus travel distances is around 1 km. The peak commuting distances for all nine transfer behaviors are greater than those of typical trips and are distributed within a range of 20~40 km. In summary, our method for extracting and analyzing transfer behaviors offers a robust and effective tool for urban transportation research, urban vitality assessment, public transportation planning, and urban planning.

The movement of people within urban areas serves as a driving force for the development of social phenomena. Accurate Origin-Destination (OD) flow data record spatial interaction patterns of individuals, goods, or information from their starting points (Origin [O]) to their destinations (Destination [D]). Precise prediction of internal city OD flows is crucial for optimizing urban traffic operational efficiency, enhancing resource utilization, and fostering sustainable urban development. However, obtaining high-quality OD flow data is challenging due to issues such as privacy protection. There are significant hurdles, including high acquisition costs, limited coverage within large areas, and sparse spatial distribution, which hinder extensive research in urban computation. Current research often relies on a single scale, utilizing extensive historical traffic data between geographic locations to predict future flows. Yet, there has been limited exploration into crucial features and model accuracy for different spatial scales. This study addresses this gap by employing taxi trajectories in Beijing and leveraging the Deep Gravity model to predict OD flow at different spatial scales. Additionally, the integration of SHapley Additive exPlanations (SHAP) values sheds light on the pivotal features influencing OD flow predictions across diverse scales. Results show that: 1) Compared to Gravity model and Radiation model, the Deep Gravity model at the street scale exhibits the highest accuracy in predicting OD flows, achieving an impressive Common Part of Commuters (CPC) value of 0.83. The Deep Gravity model effectively captures the overall structure of the OD flow network during peak morning and evening hours in Beijing, revealing a distinctive "circular dispersal" pattern; 2) For the selected spatial scales, the four features with the most significant impact on OD flow prediction accuracy are the travel distance between O and D points, the number of businesses around O and D points, the quantity of dining establishments, and the number of shopping services; 3) The local impact of the same feature on OD flow prediction models differs from its global impact. For instance, features related to education, science, and culture, as well as sports and leisure Points of Interest (POI), exhibit relatively minor effects on the model at a global scale. However, on a local scale, these features demonstrate a significant influence. This study has achieved high-precision prediction of OD flows at various spatial scales. Additionally, it quantitatively reveals the crucial factors influencing OD flow modeling at different spatial scales, thereby providing valuable insights into understanding population movements within urban areas. The findings of this research hold significant practical implications for urban planning, traffic management, and the development of smart cities.

Traditional settlements have gathered a wealth of traditional cultural resources such as ancient architecture and folklore, which have attracted significant attention for their outstanding historical, cultural and artistic values, and it is of positive significance to extract their abundant historical and cultural information and serve them for modern industrial development. Currently, there is a lack of knowledge extraction, organization and expression of the rich historical and cultural information of traditional settlements based on geographic knowledge extraction and expression perspectives to achieve the transformation of "data-information-knowledge-wisdom", this paper proposes the geographic ontology of cultural landscape genes of traditional settlements (GeoOnto-CLGTS) and explores the intrinsic correlation characteristics of the traditional landscape genes of traditional settlements. Firstly, combining the geographic information ontology and characteristics of traditional settlement landscape genes, the concept and expression method of GeoOnto-CLGTS are analyzed, and this paper proposes the construction method of GeoOnto-CLGTS model. Secondly, combing the landscape gene concepts, association relationships and data attribute characteristics, the seven-step geographic information ontology modeling method is applied to construct the conceptual layer of GeoOnto-CLGTS from top-down. By utilizing Protege tool to supplement examples using 123 traditional Chinese settlements as cases, the instance layer construction of the GeoOnto-CLGTS model is achieved. Finally, the GeoOnto-CLGTS data is stored through the Neo4j graph database to complete the construction of the knowledge graph of traditional settlement landscape genes, enabling the retrieval of landscape gene information. The results show that the GeoOnto-CLGTS constructed in this paper can provide a valuable reference for carrying out knowledge discovery of traditional settlement cultural resources and promoting digital preservation of traditional settlements in the future.

Cellular Automata (CA) provides an important tool for land use/land cover change simulation. However, previous CA models based on pure cells ignore the mixed land cover structure within cells, making it difficult to simulate the continuous evolution of mixed land systems during rapid urbanization. The Mixed-Cell Cellular Automata (MCCA) can address this issue, but its widespread application is hindered by the difficulty in obtaining fine-scale mixed structure data. To solve these problems, this study proposes a simulation analysis framework that couples the mixed pixel decomposition method with the MCCA model. This framework uses the mixed pixel decomposition algorithm to directly obtain the sub-pixel scale mixed structure data required by the MCCA model from Landsat images. The SHAP method is utilized to explore the driving forces of sub-pixel scale land cover change. To verify the proposed framework, we conducts an experiment in Wuhan city. Results show that: 1) The decomposition accuracy of the land cover data is above 0.8, and the mcFoM index of the simulation results is 0.38, indicating that this coupled model has high accuracy in characterizing the spatial pattern of mixed land structures and simulating future changes; 2) The proposed coupling model can effectively simulate the fine-scale dynamic changes of land cover proportions and discover relevant patterns of regional land use changes. For example, future land cover structure changes will mainly concentrate in built-up areas, and land mixture will experience a process of increasing first and then decreasing. Socio-economic factors such as proximity to companies, the municipal government, and high population and GDP are important driving factors for the expansion of impervious surfaces, and impervious surfaces in urban centers relatively far from high-speed railway stations grow more rapidly; 3) The future land cover change trends simulated by the proposed model are consistent with the future planning layout of Wuhan. The comparison between multiple scenarios demonstrates the MCCA model’s ability to accurately capture the subtle differences in land cover proportion between pixels. This method couples the mixed pixel decomposition method from the field of remote sensing with the mixed Cellular Automata (CA) model from the field of GIS, solving the problem of lacking fine-scale data sources for simulating mixed land cover structures. It simulates future changes in mixed land cover structures at the sub-pixel scale, which can enrich existing research on mixed land structures and provide a certain theoretical basis for urban development decisions. Additionally, it opens up new avenues for the application of CA models in other areas.

Points of Interest(POI), which are rich in semantic information, reflect current situations, and indicate areas of interest, serve as the primary data source in studies related to urban functionalization studies. These studies aim to deepen the understanding of human activities and environmental features within geographical spaces. An important research issue for enhancing the understanding of the human-environment system is detecting outliers, namely elements considerably different from the rest in large-scale spatial data. The detection of POI outliers can be broadly discussed from three perspectives: (1) spatial distribution differences, (2) spatial contextual differences, and (3) variations in the usage frequency of some POI instances and their surrounding points in specific areas due to factors such as special events, changes in urban population behavior, cultural activities, etc., leading to outliers. This paper focuses on discussing the phenomenon of POI outliers caused by spatial distribution differences. However, current outlier detection methods face with challenges. They fall short of adequately expressing and quantifying POIs' local spatial distribution features. The effectiveness of these methods needs further investigation. Given these considerations, this study proposed a novel approach for detecting POI outliers based on determination of local aggregation scales. Initially, we constructed spatial adjacency relationships of the POIs using Delaunay triangulation. Subsequently, the local aggregation characteristic scales of these points were determined by combining cross K-nearest distances and multi-scale feature parameters. Thereafter, based on the scale constraint, the points and their adjacent edge sets that met the conditions were extracted. Finally, we employed the edge length constraint index to systematically remove local long edges that did not meet the prescribed criteria. This meticulous process ensured the integration of the refined point set, thus facilitating the comprehensive detection of outliers within the POI context. The comparative experimental results, drawn from real-world data, suggested that the proposed method possessed a strong generalization ability. Moreover, it effectively and robustly detected outliers without compromising the inherent distribution characteristics of POI. We also performed an interpretability analysis of outlier detection results. The analysis revealed a close correlation between the causes of outlier distribution and various factors including the proportion of POI types, spatial layout, footprint area, and public awareness level. This study provides novel methodologies and academic perspectives for a comprehensive understanding of urban development trends, optimal resource allocation, and the enhancement of urban sustainability and quality of life.

The dense point cloud of the urban scene reconstructed by Multi-View Stereo reconstruction technology (MVS) often contains noise, resulting in surface distortion of the generated model and loss of some edge features, which cannot well reflect the real information of the reconstructed target. To solve these problems, a variational method combining 3D edge constraints is proposed to optimize the mesh model. Based on the initial grid data obtained by MVS algorithm, the energy function is constructed by the variational principle, and the grid model optimization problem is transformed into an energy function minimization problem. Firstly, the initial grid model is reconstructed from the dense point cloud. Then, the energy function is constructed by using the luminosity consistency measure, using the vertex curvature as the smooth term, and using the three-dimensional edge point constraint as the additional constraint term. Finally, the gradient descent method is used to solve the minimum energy function iteratively, and the grid deformation is driven by discretizing the gradient change to the vertex of the triangle to optimize the model. In order to construct 3D edge constraints, 3D edges must be extracted first. In this paper, 2D edges are extracted from multi-view images first, and the 2D edges are represented as multi-segment lines according to the polar constraints. Then, the 2D multi-segment line nodes are restored as 3D edge points according to the polar constraints, and the 3D edge points of the recovery points are a series of 3D multi-segment lines representing the edge outline. Finally, the edge region of the mesh model is located by taking the vertex of the mesh model closest to the 3D edge point as the neighborhood point. In this way, 3D edge features are constructed. In order to verify the effectiveness of the proposed algorithm, two real outdoor scenes from the Strecha dataset and one real indoor scene from the ETH3D dataset are selected to evaluate the reconstruction results of the proposed algorithm. In addition, the efficiency of this algorithm is analyzed by comparisons with other algorithms. Experimental results show that the proposed algorithm can effectively improve the accuracy and integrity of the grid model and retain the edge features of the target better on the grid model.

The escalating urbanization in China has exacerbated waterlogging disasters, posing substantial threats to both human lives and property. In response to the challenges of inadequate mapping and redundant map data in urban waterlogging contexts, this study introduces a comprehensive four-stage methodology for We-Map cartography. This cartography encompasses data acquisition, extraction of waterlogging points, route optimization, and scene application. The initial step involves the retrieval of social media text data through queries to the Weibo Application Programming Interface (API) within a defined timeframe. The retrieved data are subsequently subjected to thorough cleaning and preprocessing procedures. Following this, the BiLSTM-CRF model is harnessed to discern urban waterlogging locations from the social media content, thereby enhancing recognition accuracy through contextual insights. Then, users are provided with optimal route for bypassing perilous road segments, achieved via the shortest path algorithm. Leveraging the online map as the foundational framework, the We-Map is generated within the urban waterlogging setting by overlaying multiple layers. Notably, the proposed method attains an impressive overall accuracy rate of 92.7% in pinpointing urban waterlogging locations, thereby substantially enhancing mapping efficiency. A comparative analysis between map-derived waterlogging points and official records reveals a substantial overlap, thus offering valuable supplemental information to conventional monitoring techniques. Furthermore, a road network-level map of urban waterlogging points is also generated to avoid redundancies in vast geospatial information. The identified flood-prone road sections can serve as a reference, while real-time display of urban waterlogging points, coupled with the shortest path algorithm, aids in recommending optimal routes. By leveraging the inherent attributes of "we-content" within the We-Map, this method expedites rapid mapping and fulfills the exigencies of swift mapping during emergencies. To cater to diverse user needs, urban flooding scenarios map are categorized with different tags aligned with their intended applications, encompassing home-bound routes, rescue maps, driving maps, walking maps, storm assistance maps, nearest rescue supplies maps, and more. Each map is endowed with at least one tag, streamlining accurate searches and usage by other users, and concurrently providing a reference for urban rescue operations. The proposed method ensures the coherence of map content and user requisites, facilitating efficient map sharing among users. The real-time dissemination of urban waterlogging information empowers users to swiftly comprehend disaster scenes, engendering their active involvement in We-Map production, and combining optimal path recommendation to augment cartographic responsiveness in emergency disaster scenarios. This approach bears substantial practical significance and promising application potential, constituting a robust for urban waterlogging emergency responses.

With the continuous advancement of the globalization process, communication and cooperation among countries and regions around the world are becoming increasingly closer, and the scale of international migration flows is also expanding. Asia stands out as an active region for international migration, with a large portion of migratory movements occurring within its borders. In addition to the social and economic factors of the origin and destination regions, spatial and temporal dependence among migration flows is crucial in understanding international migration dynamics, indicating that migration is influenced by neighboring and past migration flows. Different from other kinds of data (e.g., regional GDP), migration flows between different regions often contain many zero values, necessitating specific methods for handling them. Additionally, spatial and temporal dependence among migration flows can be categorized into space-time contemporaneous and lagged structures, with the former reflecting the links to the preceding location and the instantaneous neighboring locations, and the latter pertaining to the preceding location and the preceding neighboring locations. Based on the bilateral migration data of Asian countries in six periods from 1990 to 2020, this study utilizes eigenvector space-time filtering models, along with contemporaneous and lagged dependent structures, as well as eigenvector spatial filtering models and zero-inflated negative binomial regression models, to explore the influential factors of the international migration flows within Asia and their changes during 1990-2020. Finally, this study aims to forecast international flows within Asia between 2020 and 2025 based on two types of space-time filtering models. Preliminary results indicate significant space-time autocorrelation of international migration flows within Asia, with neighboring migration flows exerting a greater influence over the same time period compared to the past. Incorporating eigenvectors to represent spatial and temporal dependence effectively improves the goodness-of-fit of the models. Main factors affecting international migration flows within Asia include population size, economic level, war situation, and proximity. During the 30 years (1990-2020), the influence of population size fluctuated, economic disparities initially increased before weakening, wars continued to drive emigration, geographical barriers decreased, and factors like language proximity and economic cooperation significantly influenced migration. Looking ahead from 2020 to 2025, migration trends are evident between Pakistan and India, as well as from India to Saudi Arabia, from Pakistan to Afghanistan and from Syria to Jordan. Combining the forecasting results of the two eigenvector space-time filtering models, the mean value of the total volume of international migration flows within Asia from 2020 to 2025 is projected to be approximately 1.8×10⁷. India emerges as a major country for international migration. Understanding the spatial and temporal dependence and other characteristics of international migration within Asia is crucial for accurately forecasting future migration flows and providing scientific reference for policy making.

Due to technical gaps in using satellite carbon observations for regional emission reduction, high-resolution, global-scale Fossil Fuel Carbon Dioxide (FFCO₂) emission inventories have become the main data sources for regional FFCO₂ emission research. However, there are still significant uncertainties in use of existing global-scale FFCO₂ emission inventories for regional research. Therefore, this paper quantitatively analyzed the differences and variabilities of high-resolution FFCO₂ emission inventories (ODIAC 2020b, EDGAR v6.0, and PKU-CO₂-v2) at the regional scale and fused these three inventories based on Kalman filtering algorithm. Then, this paper explored the spatial and temporal evolution pattern of FFCO₂ emissions in the Guangdong-Hong Kong-Macao Greater Bay Area (GBA). The results show that: (1) There were significant differences and variability among the current FFCO₂ emission inventories. Taking the GBA as an example, under the optimal representation spatial resolution of 3 km × 3 km, the average difference of grid cells within the region reached 140%, and the coefficient of variation was 16.3%. The use of a single global scale FFCO₂ emission inventory data for regional or urban FFCO₂ emission studies resulted in inaccurate results; (2) The reconstructed long term data from 2000 to 2018 using Kalman filter showed that the uncertainty decreased from ±15%~20% to ±10%; (3) From 2000 to 2018, the overall pattern of FFCO₂ emissions in the GBA was characterized by high emissions in Guangzhou, Shenzhen, Hong Kong, and Macao, low emission areas in the peripheral areas, and an emission transfer path from Shenzhen, Hong Kong → Guangzhou → Foshan, Dongguan → Zhongshan. The approach for regional FFCO₂ emissions proposed in this paper is demonstrated in the GBA and is applicable to other regions and cities. The conclusions of this research will provide a scientific basis for the optimal layout of energy and resources in the Greater Bay Area, which is of great significance for low-carbon transformation, high-quality development, and the construction of Beautiful Bay Area.

With the increasing improvement of the living standard of the residents in urban areas and their pursuit of quality of life, urban green spaces have become the main places of leisure and recreation for residents. Under this background, how to fairly evaluate the rationality of the layout of urban green spaces and put forward suggestions for improvement has become an important part of urban transportation and land use planning. Urban green space accessibility is a key indicator for evaluating the layout of urban green spaces. In response to the limitations of assessing attractiveness based solely on urban green space area in the past, this paper takes Suzhou urban area as an example. In addition to calculating accessibility using objective attributes in the traditional framework, the paper delves into social media data to incorporate urban residents' subjective sentiment towards urban green space quality indicators into the consideration scope of attractiveness. Through this innovative integration, the paper improves the Two-Step Floating Catchment Area (2SFCA) method, analyzing in-depth the accessibility of urban residents to urban green spaces and the dynamic changes in accessibility before and after public health emergencies. The improved 2SFCA method, combined with Sentiment Knowledge Enhanced Pre-training (SKEP) model, incorporates residents' emotional evaluations of urban green spaces to measure their subjective attractiveness. Meanwhile, considering the skewness characteristic of area indicators, the paper innovatively proposes the Scale Index (SI) as an objective attractiveness evaluation indicator for urban green spaces, providing more scientific and robust support for urban green space planning. The research findings reveal that during public health emergencies, individuals tend to prefer urban green spaces that offer convenient access, such as community parks. However, as daily life gradually resumes, there is a greater preference for urban green spaces equipped with high-quality facilities, such as specialized parks. Only considering objective area as the attractiveness of urban green space leads to overestimation of the accessibility of large-area and underestimation of small-area urban green space. Moreover, solely based on visitors' subjective quality perception of urban green space may underestimate the accessibility of communities around large urban green spaces. The improved 2SFCA method, considering both visitors' subjective perception and objective attributes of urban green space attractiveness, can more accurately assess urban green space accessibility, broadening the perspective of traditional urban green space accessibility assessment. This method can not only be applied to urban green space planning, but also provides a new idea and computational framework for the accessibility analysis of public service facilities.

As important methods for geographic information retrievals, direction relation queries have been widely applied in many fields such as data mining, intelligent reasoning, map navigation, and multi-scale data processing. In direction relation queries, it is necessary to use the direction relation models to calculate the direction relations between spatial objects. Among the proposed direction relation models, the cone-based model and the matrix model are mainly used for direction relation queries due to their simplicity and strong query capabilities. However, these two models ignore the influences of the sizes and shapes of spatial objects and the distance between them on direction relations, potentially leading to unreasonable query results. To solve the problem, this paper proposes a direction relation model that combines the cone-based model, the matrix model, and the Voronoi-based model to determine direction relations. The idea is to divide direction relations into external direction relations and internal direction relations and integrate the advantages of different models. The cone-based model and the matrix model are combined to achieve the external direction relation queries, taking into account the influences of the sizes of spatial objects and the distance between them on direction relations. The Voronoi-based model is employed for the internal direction relation queries, considering the influences of the shapes of spatial objects on direction relations. The experimental results show that the combinational model has good applicability and feasibility in direction relation queries, maintaining high consistency with people's spatial cognitions. The main advantages of the combinational model are that: (1) it fully considers the influences of the sizes and shapes of spatial objects and the distance between them on direction relations, and overcomes the disadvantages of the cone-based model and the matrix model in direction relation queries; and (2) it integrates the strengths of the cone-based model, the matrix model, and the Voronoi-based model, enabling the unified querying of external direction relations and internal direction relations and resulting in improved accuracy of direction relation queries. Moreover, the combinational model will help improve the accuracy and reliability of spatial data processing such as intelligent querying and reasoning of spatial information and the calculation of multi-scale spatial relation similarity.

Accurate and explainable prediction of PM_2.5 concentration can help humans avoid exposure risks to air pollution, which is of great significance for human health risk assessment and policy implementation. Currently, the existing PM_2.5 concentration prediction models focus on improving the model prediction accuracy without considering model interpretability, resulting in poor model reusability and trustworthiness. Therefore, this paper proposes an Attentional Spatiotemporal Ordinary Differential Equation (ASTODE) model for PM_2.5 concentration prediction tasks considering both prediction accuracy and model interpretability. Specifically, this paper integrates the Neural Ordinary Differential Equation (NODE) into the PM_2.5 concentration prediction task to improve the interpretability of the prediction model. In addition, to address the challenge of traditional NODE in mining spatial dependencies in PM_2.5 concentration data, this paper proposes a novel spatiotemporal derivative network that extends the traditional NODE to spatiotemporal ordinary differential equations. To address the challenges of traditional NODE in mining long-term dependencies in PM_2.5 concentration data, this paper proposes a spatiotemporal attention mechanism to fuse hidden states of multiple time nodes. In the experimental section, the proposed ASTODE model is validated using a real PM_2.5 concentration dataset. This paper quantifies the prediction errors of the ASTODE model in both temporal and spatial dimensions. By comparing with six existing baseline methods, our proposed ASTODE model obtains a similar or higher prediction accuracy. This paper also analyzes the interpretability of our proposed ASTODE model from a visualization perspective, demonstrating that the proposed ASTODE model balances the prediction accuracy and interpretability to some extent.

With the acceleration of population aging, the urban built environment for the elderly faces severe challenges. Urban street environments, one of the most frequently used places by the elderly, require high-quality construction, which is vital for realizing an age-friendly society. However, few studies have focused on the spatial effects and influencing factors of urban street environment quality for the elderly from a large-scale and human perspective, resulting in difficult practical applications. Therefore, this study took Tianhe district, Guangzhou as a study area, using machine learning and deep learning technology to evaluate the urban street environment quality for the elderly and analyze its spatial distribution and influencing mechanisms. Based on 14 916 human-centric street view images taken by panoramic cameras, semantic segmentation and object detection techniques were used to extract environmental elements. Greenness, openness, crowdedness, enclosure, sidewalk ratio, and scene diversity were obtained finally as explanatory variables in this study. A human-machine adversarial scoring system was constructed for the age-friendly street environment quality assessment. Twenty-two elderly volunteers were invited to rate their sense of walkability, vitality, security, belonging, and pleasure from 1 000 randomly selected images. A residual neural network 50 (ResNet50) was used to predict the urban street environment quality in the Tianhe district based on street view images and crowd-sourced data. The spatial autocorrelation was measured by global and Local Moran's I. Ordinary Least Square regression model (OLS), Spatial Lag Model (SLM), and Spatial Error Model (SEM) were established to analyze the influence mechanisms. Results show that: (1) Using human-centric street view images, machine learning, and spatial statistics methods, this study conducted a fast, large-scale, and precise age-friendly street environment quality assessment and accounted for spatial heterogeneity to identify its key influencing factors; (2) There was a moderate degree of spatial aggregation of different street environment qualities for the elderly in the Tianhe district. For older people, commercial streets and streets near low-density residential areas were associated with higher levels of walkability, activity, sense of safety, and pleasure. Although waterfront streets had higher levels of activity and security, the level of pleasure was low. Streets near high-density residential areas were found to have lower levels of activity level, sense of safety, and pleasure. The sense of belonging was higher in commercial streets and lower in streets close to residential areas; (3) The effects of environmental factors on different street environment quality indexes for the elderly were different. Greenness, openness, and enclosure were important factors while visual crowdedness, sidewalks, and scene diversity played a weak role. Greenness had a positive effect on activity level and sense of safety, but a negative effect on pleasure and sense of belonging. Openness was positively correlated with walkability, pleasure, and sense of belonging, and negatively correlated with activity levels. Enclosure had negative effects on all indicators, especially the sense of belonging. These results reveal the spatial association, heterogeneity, and influencing mechanisms of the street environment quality for the elderly based on human-centric street view images, machine learning, and deep learning techniques on a large urban scale. It shows a feasible paradigm to analyze the street environment for the elderly, providing practical implications to build resilient streets more conducive to an age-friendly society. It's of great value for policy-making, urban planning, and management.

Visualization of uncertainty is a research focus and difficulty in the field of cyberspace map visualization. Reasonable design of uncertainty symbols is crucial for the quick reading, mining, accurate analysis, and decision making of cyberspace map information. In this paper, a symbolic representation of double variables uncertainty in cyberspace based on data model of multi-granularity spatiotemporal objects is proposed. This method can solve the problem that the variable symbols in the cyberspace node link graph cannot reflect the expression uncertainty of nodes and connected edges in a timely and efficient manner. Taking geographic social networks as an example, we first adopt the modeling method of multi-granularity spatiotemporal objects and divide the cyberspace into carrier class, subject class, and data class based on the four classifications of cyberspace proposed by Academician Fang Binxing. The entity is divided specifically from the virtual and real perspectives. Then we analyze the content and process of cyberspace object modeling, design cyberspace entity object class, and create spatiotemporal objects. Based on this, combining the problems of traditional symbols in the expression of cyberspace node link graph, the uncertainty expression theory and uncertainty expression model of cyberspace are analyzed. The expression theory reveals that the uncertainty in cyberspace generated from Single Variables expression is larger than that from Double Variables expression over time. The uncertainty expression model divides the uncertainty expression of object data in cyberspace into node uncertainty, edge uncertainty, local uncertainty, and global uncertainty. Then the vizent symbols of nodes and edges of 1-8 levels are made respectively. Finally, a case study is carried out. Firstly, the presentation of application results is conducted. The experiment first constructs the cyberspace object, instantiates it, and then visualizes it with symbols. Secondly, a symbolic control experiment is carried out. The control experiments are carried out from four categories: primary and secondary values, width brightness, saturation transparency, and vizent symbols of the experimental group. The results of the symbol experiment are tested by statistical methods. The results show that the method based on objectified modeling is conducive to the expression of multi-granularity, all-type, and multi-dimensional dynamics of cyberspace, and the development and change of cyberspace can be vividly, intuitively, and comprehensively expressed through visualization and interaction technology. The newly designed vizent symbol has a good effect on the expression of double variables uncertainty difference in cyberspace, which is helpful to obtain the uncertainty information in cyberspace timely, efficiently, and accurately. This study provides references for the development of the field of map visualization in cyberspace.

With the development of 3D sensors and 3D reconstruction techniques, the registration and fusion of cross-source point clouds have become a research hotspot. However, traditional registration methods use a single feature as the registration primitive, which leads to problems such as weak spatial geometric constraints. Combining multiple structural features with joint constraints can improve the registration accuracy to a certain extent. In order to fuse cross-source point cloud data with high accuracy and fully express the façade information in the scene, this paper proposes a cross-source point cloud registration method based on the constraints of line and surface features. Firstly, the homonymous line and plane features in the cross-source point cloud are extracted by RANSAC algorithm, which are mainly used to constrain the point cloud model in registration. Then the quaternion method is used to describe the spatial transformation parameters based on the line and surface features. The rotation and transformation in arbitrary 3D space can be realized at a faster calculation speed compared with other representations while also avoiding the gimbal lock phenomenon. The line features are used as the constraints of registration, the spatial transformation objective function is constructed, and the parameters related to the transformation are estimated to complete a coarse registration and solve the scale variability. Based on the coarse registration, the surface features are further used as the constraints to solve the rotation matrix and translation parameters to achieve a fine registration. The use of surface features instead of point features as the registration primitives can avoid the selection of common points from massive point cloud data, reducing the accidental errors selected by human selection, avoiding the accumulation of errors, and further improving the registration accuracy. Finally, experiments are conducted using the image-matched point clouds and LiDAR point cloud data for a small area and a large area to study the feasibility of this paper's method in different scales. Results show that the RMSE values for the small-area single building, multiple buildings, and large-area building clusters are 0.364 7, 0.032, and 0.614 6, respectively. The maximum angle between the homonymous surfaces does not exceed 1.5°, the minimum is less than 0.1°, and the mean value of the angle is within the range of 1°. The coarse registration based on line feature constraints can solve the scaling problem well in different scenarios, and the fine registration based on surface feature constraints can further improve the accuracy of the rotation matrix and translation parameters. These results indicate that the cross-source point cloud registration method based on line and surface feature constraints is feasible at different scales.

The coastline is located at land-sea interaction zone, and its utilization pattern is closely related to the spatial planning on both sides of the coastline (i.e., land and sea). This study primarily used the 2022 SDGSAT-1 satellite imagery and considered both the land and sea sides of the coastline to investigate the correlation characteristics of mainland coastlines and the spatial utilization on land and sea sides in China from the three levels: the entire coastal zone, the provincial (municipal) coastal zones, and the profile lines. The results show that: (1) The overall utilization ratio of the mainland coastline was high, with only 35.11% remaining undeveloped. Among the developed coastline types, aquaculture embankment coastlines accounted for the highest proportion, followed by port and dock coastlines; (2) In terms of land use, the largest area was utilized for construction purposes, with a significant proportion in Tianjin and Shanghai, followed by the artificial wetland area, with a high proportion in Liaoning and Shandong provinces. In terms of sea area use, fishery dominated in Shandong and Fujian provinces, marine transportation was most prominent in Shanghai, and the energy development accounted for a large proportion in Shanghai, Tianjin, and Hebei; (3) Regarding utilization intensity, there were clear positive correlations between the Index of Coastline Utilization Degree (ICUD), Index of Land Utilization Degree (ILUD), and Index of Sea Utilization Degree (ISUD). Especially, the correlation between coastline utilization and land use activities was relatively strong. Spatially, the regions with high and low values for the three indices were generally consistent. High-value regions were mainly distributed in northern regions such as Tianjin and Hebei, while low-value regions were predominantly found in southern provinces such as Zhejiang and Fujian; (4) Observing both land and sea sides from the coastline perspective, the landward side of unused coastlines was mostly covered by forests, with a lowest index of land utilization degree, while on the seaward side of unused coastlines, the proportion of tourism and recreational sea usage was relatively high. Port and dock coastlines exhibited the highest comprehensive indices on both land and sea sides. Construction lands dominated the landward side and energy development and transportation dominated the seaward side. The proportion of ecological protection sea usage on the sea side of aquaculture and salt embankment coastlines was relatively high. These findings provide scientific references and decision support for the scientific planning and management of coastlines, as well as the optimization of land-sea spatial structures.

Accurate matching of line features is of paramount importance in the reconstruction and optimization of three-dimensional models. However, traditional dual-view line matching encounters challenges due to a limited number of views, resulting in suboptimal robustness in line matching. For line extraction results with breaks, the number of lines extracted for the same line on different images is different, resulting in poor integrity of straight line matching results. To address these issues, this paper proposes a multi-view line matching algorithm that combines Multiple-View Stereo (MVS) and Leiden graph clustering. The algorithm commences by employing the line extraction algorithm and the MVS three-dimensional reconstruction algorithm on input multi-view images for line information extraction and multi-view three-dimensional information extraction, respectively. This process yields lines on each view, dense three-dimensional points encapsulating the image scene, and the correspondence between object-side three-dimensional points and their corresponding image-side two-dimensional points. Building upon this foundation, the algorithm constructs line descriptors in the image domain by considering lines and their matching point sets within their neighborhoods. Subsequently, leveraging the three-dimensional line projection angle constraints, point-line position relationship constraints, and corresponding point constraints, the algorithm filters matching candidates based on these three geometric constraints. Harnessing the similarity relationships between lines on each view, an undirected graph is constructed. Here, lines on each view serve as nodes, and the similarity scores between lines act as edge weights. Simultaneously, connected components composed of single nodes are removed from the undirected graph, resulting in the set of connected components that represent the initial matching results. In the final stage of this process, nodes of each connected component are reconnected based on same-view collinear constraints, forming many sub-undirected graphs. The Leiden algorithm is then applied to cluster the nodes of these sub-undirected graphs. The clusters composed of a single node in the clustering results represent unsuccessfully matched lines, while clusters composed of two or more nodes signify the presence of corresponding lines across multiple views. Ultimately, the algorithm achieves accurate line matching on multi-view images. The experimental results show that the line matching results using the proposed algorithm are improved in terms of the number of line matches and the matching accuracy relative to other comparison algorithms.

Disaster early warning plays an important role in disaster reduction management by proactively disseminating disaster information to guide residents in taking timely evacuation actions, thus effectively reducing disaster losses and casualties. The dynamic response process of residents to disaster early warning information and the assessment of the effectiveness of different flood disaster early warning strategies are pressing issues. This paper proposes a simulation method for urban rainstorm flood disaster early warning strategies based on Agent-Based Modeling (ABM). Firstly, three warning strategies are established: rainfall forecast-based, flood inundation-based, and population exposure-based. Secondly, individual risk perception is assessed by considering a variety of socio-demographic characteristics, and a probabilistic model of individual travel decision-making is constructed. Based on this, an agent-based model for urban flood disaster early warning strategies is developed. Finally, taking Futian District in Shenzhen, China as a case study, residents' travel behavior and flood risk are simulated and analyzed with different flood warning strategies under 20-year, 50-year, and 100-year return period rainstorm scenarios. The results show that: (1) The ABM simulation model, considering residents' perception of flood disaster risk and the probability of individual travel decision-making, accurately simulates residents' travel response behavior and changes in flood disaster risk under different warning strategies. It provides a scientific and comprehensive evaluation of the effectiveness of urban flood disaster early warning strategies; (2) Different warning strategies lead to significant differences in population travel response behavior, resulting in varying effectiveness in reducing urban rainstorm flood disaster risk. Faced with a 20-year rainfall scenario, flood inundation-based and population exposure-based early warning strategies help residents in the study area quickly identify high-risk areas, significantly reducing the risk to buildings and roads. Faced with a 20-year return period rainstorm scenario, the study area shows minimal changes in residents' travel behavior under rainfall forecast-based warnings. However, flood inundation-based, and population exposure-based warning strategies help residents rapidly identify high-risk areas, significantly reducing the number of people heading to red and orange warning zones. This results in a noticeable decrease in risks to buildings and roads; (3) Under different rainstorm scenarios, the effectiveness of various flood disaster early warning strategies varies. In the face of smaller rainstorm scenarios, refined flood disaster early warning strategies, such as flood inundation-based, and population exposure-based, demonstrate effectiveness in reducing urban flood disaster risk. However, when dealing with extreme rainstorm scenarios, adopting a unified flood disaster early warning strategy, such as rainfall forecast-based, is more effective than a refined warning strategy. Therefore, urban flood disaster early warning systems should be tailored to local conditions and varying circumstances, establishing a graded, zonal, and scenario-based warning system.

Image super-resolution technology enhances image clarity and enriches image detail by improving image spatial resolution, enabling quality enhancement without changing hardware conditions. Given the large size, complex target features, and abundant details of remote sensing images, along with the need for efficient information acquisition, we propose a Diffusion Super-Resolution (DSR) algorithm based on a conditional diffusion model. This approach uses low-resolution remote sensing images from the same region as conditioning inputs to the diffusion model, while high-resolution images with added noise are concatenated as inputs. A deep noise training network was constructed with U-Net as the backbone, incorporating residual connections and self-attention mechanisms. The loss function was also improved for better super-resolution results. The DSR method was tested using high-resolution remote sensing images from multiple periods of the domestic Gaofen and SuperView satellite series. The super-resolution results demonstrated pixel dimension expansion from 32 to 128. Comparative experiments with Bicubic, SRGAN, Real-ESRGAN, and SwinIR super-resolution algorithms showed that the DSR method outperforms these algorithms in both PSNR and SSIM metrics. Additionally, the DSR method significantly improves the quality of multispectral remote sensing images. By leveraging the conditional diffusion model, it successfully preserves rich detail and enhances spatial resolution without compromising image clarity. This method offers an efficient solution for super-resolution reconstruction, ensuring effective information acquisition in remote sensing applications and fulfilling the requirements of various domains such as land use classification, environmental monitoring, and urban planning. Moreover, the DSR method also opens new avenues for future research by demonstrating the potential of diffusion models in remote sensing image processing. It overcomes the limitations of simple convolutional networks, which extract only shallow features, and avoids the convergence issues commonly seen in adversarial neural networks during training, ultimately improving the restoration of rich details in remote sensing images.

Large outdoor point clouds have rich spatial structures and are one of the important means of obtaining geographic information. They have broad application prospects in fields such as autonomous driving, robot navigation, and 3D reconstruction. Due to its inherent irregularity, complex geometric structural features, and significant changes in land scale, the accuracy of point cloud segmentation remains a huge challenge. At present, most point cloud segmentation methods only extract features based on the original 3D coordinates and color information of point cloud data and have not fully explored the information contained in point cloud data with rich spatial information, especially the problem of insufficient utilization of geometric and color information in large-scale point clouds. In order to effectively address the aforementioned issues, this paper introduces the CMGF-Net, a method for semantic segmentation of point clouds that effectively integrates color information and multi-scale geometric features. In this network, dedicated modules are designed for extracting geometric feature information and semantic feature information. In the geometric feature information extraction path, to fully leverage the geometric characteristics of point cloud data, two feature extraction modules are designed: the Relative Position Feature (RPF) extraction module and the Local Geometry Properties (LGP)extraction module, both focusing on the characteristics of the local neighborhood. In the RPF module, spatial normal information of the 3D point cloud and relative spatial distances are utilized to extract the relative positional relationships between neighboring points and the central point. The LGP module exploits the unique performance characteristics of point cloud geometric properties across different terrains, integrating geometric attribute features from the local region. Subsequently, the designed Local Geometric Feature Fusion module (LGF) combines the extracted feature information from the RPF and LGP modules, yielding fused geometric feature information. Furthermore, to learn multi-scale geometric features from the point cloud, CMGF-Net conducts geometric feature extraction at different scales within the network layers. Eventually, the extracted geometric features are hierarchically fused with semantically extracted features based on color information. By extracting multi-scale geometric features and integrating semantic features, the learning ability of the network is enhanced. The experimental results show that our proposed network model achieves a mean Intersection Over Union (mIoU) of 78.2% and an Overall Accuracy (OA) of 95.0% on the Semantic3D dataset, outperforming KPConv by 3.6% and 2.1%, respectively. On the SensatUrban dataset, it achieves a mIOU of 59.2% and an OA of 93.7%. These findings demonstrate that the proposed network model, CMGF-Net, yields promising results in the segmentation of large-scale outdoor point clouds.

Optical images and SAR images have rich complementary attributes, and an effective data fusion strategy can provide a solid information base for objects interpretation. Roads, as strip features, their topology, distribution patterns, and application scenarios often pose challenges to the interpretation results. Based on this, this paper proposes a multi-branch and dual-task method for road extraction from multimodal remote sensing images. First, encoding-decoding networks with the same structure but independent parameters are constructed for feature extraction of optical and SAR images, respectively, and road surface segmentation labels are used for supervised training. Second, the coding layer features of the SAR images are introduced for road edge detection, and their intermediate features are input to the decoding layer features of the SAR image, so as to optimize the discrimination effect between the road and the background. Finally, the designed Channel Attention-Strip Spatial Attention (CA-SSA) is utilized to fully fuse the shallow and deep features of optical and SAR images to predict the final road extraction results. In the experiment, using the Dongying data set as the reference, it is proved that the method of this paper is superior to the comparative methods based on quantitative evaluation metrics, has obvious advantages in challenging areas such as road intersection and low-grade roads, and has best road extraction results when optical images is affected by clouds.

[Significance] As a representative of new-quality productivity, the low-altitude economy is gradually emerging as a new engine for economic growth. This economy is based on the development and utilization of low-altitude airspace resources. While bringing development opportunities to geospatial information technology, it also poses entirely new challenges. [Progress and Analysis] In this paper, we introduce the division of low-altitude airspace resources and highlight typical drone application scenarios in the context of the low-altitude economy. Subsequently, we analyze the broad application prospects of geospatial information technology in key areas of the low-altitude economy, including the refined utilization of airspace resources, the construction of low-altitude environments, the planning, construction, and operation of new air traffic infrastructure, as well as the safe and efficient operation and regulatory oversight of drones. We emphasize that the geospatial information industry will benefit from development opportunities such as the integration and innovation of emerging scientific and technological advancements, growing market demand, policy support, industrial guidance, and industrial upgrading and transformation. [Prospect] Finally, we briefly address the challenges geospatial information technology must overcome to meet the development needs of the low-altitude economy. These include advancements in spatio-temporal dimension elevation, map and location-based services, high-frequency and rapid data acquisition systems, all-time and all-domain capabilities, and ubiquitous intelligent technologies. These areas will also serve as future directions for development and breakthroughs in geospatial information technology.

Technological knowledge has become the key element of regional innovation and development in the new era. Exploring the inherent mechanism of the growth and development of technology transfer network is of great significance to improve the vitality of regional innovation. However, the endogenous mechanisms and spatial differences of technology transfer network evolution is rarely studied. This study aims to gain a better understanding of the growth and development process of urban technology transfer networks in China and their spatial differences. First, this paper takes 282 cities of China as research units. Second, information on patent transferred data is subjected to ownership linkage mode to construct the urban technology transfer network, resulting in a panel dataset of 282 cities in China in 2001—2020. Finally, stochastic actor-oriented models for the evolution of networks are constructed to study the evolution of technology transfer networks and spatial heterogeneity. Results show that: (1) The evolution pattern of urban technology transfer network in China presents a "core-periphery" structure. The network exhibits strong polarization characteristics, but it is decreasing gradually. The increasingly complex tripartite relationship between cities is an important feature of network evolution. These tripartite relations not only affect the formation of link relations but also promote the differentiation of local levels of the network; (2) Endogenous structural factors are the key factors for the growth and development of urban technology transfer network in China. Reciprocity and network closure constitute the micro basis of the evolution of urban technology transfer network. Path dependence is a key force in strengthening the link relationship between urban technology transfer networks; (3) The endogenous mechanism of the evolution of urban technology transfer network in China has obvious spatial heterogeneity. In the southern region of eastern China, the urban technology transfer network has strong dynamics. Reciprocity, network closure, and path dependence have become the endogenous driving forces for the growth and development of technology transfer networks. In the northern region of eastern China, the evolution rate of the network shows a downward trend. Reciprocity and path dependence contribute to the formation of network link pattern. In the northwest inland and the Qinghai-Tibet Plateau, the network evolution rate tends to increase, but the network density is small, and only the reciprocity effect is significant. This paper will deepen the understanding of the evolution law of urban networks and provide a scientific reference for China's urban innovation and development policy.

Construction waste is an inevitable byproduct of urban renewal processes, causing serious environmental pollution and ecological pressure. Precisely quantifying the annual production of urban construction waste and the resource conversion rate is crucial for assessing the cost of urban renewal. Traditional manual methods of estimating construction waste production rely heavily on statistical data and historical experience, which are inflexible, time-consuming, and labor-intensive in practical application, and need improvement in terms of accuracy and timeliness. Existing deep learning models have relatively poor capabilities in extracting and integrating small targets and multi-scale features, making it difficult to handle irregular shapes and fragmented detection areas. This paper proposes a Multi-Scale Feature Fusion and Attention-Enhanced Network (MS-FF-AENet) based on High-resolution Remote Sensing Images (HRSIs) to dynamically track and detect changes in buildings and construction waste disposal sites. This paper introduces a novel encoder-decoder structure, utilizing ResNet-101 to extract deeper features to enhance classification accuracy and effectively mitigate the gradient vanishing problem caused by increasing the depth of convolutional neural networks. The Depthwise Separable-Atrous Spatial Pyramid Pooling (DS-ASPP) with different dilation rates is constructed to address insufficient receptive fields, resolving the issue of discontinuous holes when extracting large targets. The Dual Attention Mechanism Module (DAMM) is employed to better preserve spatial details, enriching feature extraction. In the decoder, Multi-Scale Feature Fusion (MS-FF) is utilized to capture contextual information, integrating shallow and intermediate features of the backbone network, thereby enhancing extraction capabilities in complex scenes. MS-FF-AENet is employed to extract and analyze changes in building areas at different time periods, calculating the engineering waste from new constructions and demolition waste from demolished buildings, thereby obtaining the annual production of urban construction waste. Furthermore, MS-FF-AENet is utilized to extract construction waste disposal sites at different time periods, estimating the amount of construction waste landfill based on changes in landfill waste, indirectly assessing the resource conversion rate of urban construction waste. Based on HRSIs of Changping District, Beijing from 2019 to 2020, experimental results demonstrate: (1) Among a series of baseline models including UNet, SegNet, PSPNet, DeepLabV3+, DSAT-Net、ConvLSR-Net and SDSC-UNet, MS-FF-AENet exhibits advantages in terms of precision and efficiency in extracting buildings and construction waste; (2) During the period from 2019 to 2020, the annual production of construction waste in the study area due to urban renewal is approximately 4 101 156.5 tons, with approximately 2 251 855.872 tons being landfill waste and approximately 1 849, 300.628 tons being resource conversion waste, resulting in a construction waste resource conversion rate of 45.09%, further corroborating government statistical reports. This paper provides a convenient and effective analysis approach for accurate measurement of the cost of urban renewal.

The use of Unmanned Aerial Vehicles (UAVs) for road image collection is advantageous owing to their large scope and cost-effectiveness. However, the size and shape of road damages vary significantly, making them challenging to predict. Furthermore, due to the limitations of computational resources, generalized target detection algorithms are only applicable to small-size images (512 pixels× 512 pixels or 640 pixels× 640 pixels). This makes them unsuitable for direct application to large-size UAV images (5 472 pixels× 3 648 pixels or 7 952 pixels × 5 304 pixels). The utilization of traditional methods for the detection of multi-scale targets in large-size images is associated with a number of issues, including the slicing of large-size targets and the failure to detect small-size targets. To address these challenges, this paper presents an innovative solution that combines the global-local multiscale fusion strategy with YOLOv5-RDD. First, a YOLOv5-RDD model is constructed, and based on the existing YOLOv5 model, a multiscale C3 (MSC3) module and a Contextual Feature Pyramid Network (CFPN) are designed to improve the detection capability of multiscale targets. Additionally, we introduce an extra detection head for larger-size targets. Then, a global-local multiscale fusion strategy is proposed, which uses resizing and slicing means to obtain global and local information of large UAV images, and then superimposes the global and local multiscale information to obtain the multi-scale information of the whole large image. The detection results are optimized using the center non-maximum value suppression algorithm. Specifically, the global-local multiscale fusion strategy first trains the YOLOv5-RDD using multiscale training strategy to learn complete multiscale features. Then, YOLOv5-RDD predicts multiscale road damages in large-size images using a multiscale prediction strategy to avoid directly applying it to these images. Finally, we use center non-maximum suppression to eliminate redundant object detection boxes. To verify the effectiveness of the proposed method and meet real-world requirements, a UAV-RDD dataset specialized for UAV road disease detection is created. The experimental results show that compared with the original YOLOv5 model, the new model YOLOv5-RDD improves the mAP by 5.8%, while the global-local multiscale fusion strategy improves the mAP by 9.73% compared with the traditional method. The MSC3 achieves the maximum enhancement of mAP@0.5, with an improvement of 2.6%, contributing only 0.8 M parameters. The CFPN yields an improvement of 0.2% in mAP@0.5 while reducing the number of parameters by 8 M. These results fully prove the effectiveness and superiority of the method in this paper.

[Objectives] To systematically review recent advancements in text-to-image generation technology driven by large-scale AI models and explore its potential applications in urban and rural planning. [Discussion] This study provides a comprehensive review of the development of text-to-image generation technology from the perspectives of training datasets, model architectures, and evaluation methods, highlighting the key factors contributing to its success. While this technology has achieved remarkable progress in general computer science, its application in urban and rural planning remains constrained by several critical challenges. These include the lack of high-quality domain-specific data, limited controllability and reliability of generated content, and the absence of constraints informed by geoscience expertise. To address these challenges, this paper proposes several research strategies, including domain-specific data augmentation techniques, text-to-image generation models enhanced with spatial information through instruction-based extensions, and locally editable models guided by induced layouts. Furthermore, through multiple case studies, the paper demonstrates the value and potential of text-to-image generation technology in facilitating innovative practices in urban and rural planning and design. [Prospect] With continued technological advancements and interdisciplinary integration, text-to-image generation technology holds promise as a significant driver of innovation in urban and rural planning and design. It is expected to support more efficient and intelligent design practices, paving the way for groundbreaking applications in this field.

The rapid development of information and communication technologies and mobile computing has generated a variety of mobility big data, providing new opportunities for understanding and exploring the spatiotemporal distribution and mobility characteristics of resident travel, and further contributing to the construction of smart cities. However, the emerging mobile data have experienced significant growth in both scale and complexity compared to traditional data, posing challenges for its structural characteristic analysis. To address these issues, this paper proposes an analytical framework to deal with the spatiotemporal distribution characteristics of high-dimensional ride-hailing travel pattern. Compared to traditional square partitions, a regular hexagon is closer to a circle, and the six adjacent hexagons connected to its edges are symmetrically equivalent, which can be more advantageous in aggregating demands with similar travel characteristics into the same partition. Therefore, hexagonal partition is selected as the basic clustering unit, and different spatiotemporal patterns are identified by clustering homogeneous travel distribution groups. Firstly, the spatiotemporal characteristics of travel distribution aggregated in the hexagonal partition are summarized into three main components: the departure demand distribution at the origin partition, the spatial distribution at the destination partition, and the arrival demand distribution at the destination partition. The spatiotemporal similarity between two partitions can be expressed as the product of these three types of distribution similarity. Furthermore, a Clustering Algorithm with Fast Search and Find of Spatiotemporal Density Peaks (CFSFSTDP) is proposed to identify the spatiotemporal patterns of ride-hailing travel distribution in each partition. The spatiotemporal distances between different partitions are obtained through the calculation of spatiotemporal similarity. Finally, affinity propagation clustering algorithm is used to perform clustering analysis on the time series variation pattern of spatiotemporal pattern of travel distribution in each partition. The time series similarity of spatiotemporal patterns between different partitions is represented by the sum of Euclidean distances between time series of each interval, and the model converges through continuous updates of attractiveness and affiliation indices. Through the empirical analysis of Didi Chuxing order data in Chengdu for one month, the validity of the method is verified. Based on the identified seven spatiotemporal distribution patterns, the differences of spatiotemporal patterns in the size, location, and time of demand are analyzed, and the functional types of ride-hailing travel in different partitions are discussed. The identified six time series patterns better grasp the time continuity of spatiotemporal patterns of ride-hailing travel distribution and help to better build the corresponding spatiotemporal evolution digital.

High-quality precipitation data is essential to guarantee meteorological services and hydrological applications. As an important source of precipitation data, satellite precipitation data, at various spatial and temporal scales, has been widely used in the field of hydrology and meteorology. However, satellite precipitation products often suffer from issues such as coarse spatial resolution and low accuracy, hindering their suitability for refined hydrological and meteorological applications. This study uses Random Forest (RF) as the basic model and proposes a spatial Random Forest multi-source fusion (SRF-MF) method to fuse daily precipitation data.This method first uses spatial random forest to downscale various satellite precipitation products on a monthly scale, leveraging the strong correlation between precipitation and environmental factors. Then it decomposes the monthly precipitation into daily values based on daily ratios. Finally, it uses RF to fuse downscaled data and site data to ultimately generate high-quality daily precipitation data. Using five satellite precipitation products (CHIRPS, CMORPH, PERSIANN, GAMaP and IMERG), along with rain gauge data from Sichuan Province spanning from 2015 to 2019, the SRF-MF method generated a daily precipitation dataset. This dataset was compared with the original satellite precipitation and five machine learning methods, including Random Forest Merging method (RF-MEP), single machine learning method (RF, ANN), and dual machine learning method (RF-RF, RF-ANN). Experimental results demonstrate that the precipitation dataset generated by the SRF-MF method exhibits significantly higher accuracy compared to several other methods across different time scales (daily, monthly, quarterly, yearly), with greater precision in capturing precipitation events of varying intensities. Moreover, the spatial details of the precipitation distribution map are richer and more accurate. The findings of this article provide research ideas for improving the quality of satellite precipitation data and expanding its application fields.

In the context of the rapid development of urbanization, the reasonable selection of locations for public service facilities is critical for delivering efficient services and enhancing the quality of urban residents' lives. However, prevailing approaches for allocation of public service facilities often fall short of meeting the demands on their performance and efficiency in complex and large-scale real-world scenarios. To address these issues, this article proposed a novel Graph-Deep-Reinforcement-Learning Facility Location Allocation Model (GDRL-FLAM), coupling a Facility Location Allocation Graph Attention Network (FLA-GAT) with a Deep Reinforcement Learning (DRL) algorithm. This proposed model tackled the location allocation problem for public service facilities based on graph representation and the REINFORCE algorithm. To assess the performance and efficiency of the proposed model, this study conducted experiments based on randomly generated datasets with 20, 50, and 100 points. The experimental results indicated that: (1) For the tests with 20, 50, and 100 points, the GDRL-FLAM model exhibited a significant improvement ranging from 11.79% to 14.49% compared to the Genetic Algorithm (GA) which is one of the commonly used heuristic algorithms for addressing location allocation problems. For the tests with 150 and 200 points, the improvement ranged from 1.52% to 9.35%. Moreover, with the increase in the size of the training set, the model also demonstrated enhanced generalizability on large-scale datasets; (2) The GDRL-FLAM model showed strong transfer learning ability to obtain the location allocation strategies in simple scenarios and adapt them to more complex scenarios; (3) In the case study of Singapore, the GDRL-FLAM model outperformed GA significantly, achieving obvious improvements ranging from 1.01% to 10.75%; (4) In all these abovementioned tests and experiments, the GDRL-FLAM model showed substantial improvement in efficiency compared to GA. In short, this study demonstrated the potential of the proposed GDRL-FLAM model in addressing the location allocation issues for public service facilities, due to its generalization and transfer learning abilities. The proposed GDRL-FLAM could also be adapted to solve other spatial optimization problems. Finally, the article discussed the limitations of the model and outlined potential directions for future research.