Road Vector Research Articles

Large-scale integration of remotely sensed and GIS road networks: A full image-vector conflation approach based on optimization and deep learning

Road networks play an important role in the sustainable development of human society. Conventionally, there are two sources of road data acquisition: road extraction from Remote Sensing (RS) imagery and GIS based map production. Each method has its limitations. The RS road extraction methods are primarily raster-based and the extracted roads are not directly usable in GIS due to their fragmented and noisy nature, while vector-based methods cannot utilize rich raster information. Further more, the vector and raster data can have discrepancies for various reasons. Efficient road data production requires an image-vector conflation process that can match and combine raster and vector-based road data automatically.In this study, we propose a full image-vector conflation framework that directly integrates image and vector road data by appropriately transforming extracted roads from imagery and establishing a match relation between these roads and a credible target GIS road dataset. Based on analyzing these match relations, we propose new metrics for measuring the degree of agreement between the raster and vector road data. The proposed framework combines state-of-the-art deep learning methods for image segmentation and optimization-based models for object matching. We prepared a large-scale high-resolution road dataset covering two counties in Kansas, US. Using trained models from one of the two counties, we were able to extract road segments in the other county and match them to the TIGER/Line roads.Our experiments show that conventional performance metrics for road extraction (e.g. IoU) are insufficient for measuring the degree of agreement between image and vector roads as they are pixel-based and are too sensitive to spatial displacement. Instead, the newly defined vector-based agreement metrics are needed for image-vector conflation purposes. Experiments show that, by the vector-based metrics, nearly 90% of GIS road lengths in the study area were extracted and over 90% of extracted roads matched the target GIS roads. The new framework streamlines raster-vector conflation of roads and can potentially expedite relevant geospatial analyses regarding change detection, disaster monitoring and GIS data production, among others.

Mountainous road vector data update method based on matching point pair grouping

ABSTRACT The vector information of mountainous roads is great significance, including regional economics, emergency response, disaster management, among others. Nevertheless, there are irregular deformations in existing vector data, coupled with incomplete extraction results for mountainous roadways. These issues pose challenges in establishing a matching model between road extraction results and existing vector data, significantly reducing the efficiency of automated updates for the current vector data. Targeting roads as a distinct artificial feature and leveraging their topological connectivity, we propose a vector data updating method based on matching point pair grouping. Firstly, we employ the SDUNet model to extract mountainous road and utilize the Superglue model to match road binary images with vector binary images, resulting in sets of initial matching point pairs. Subsequently, we propose a road line extraction model that groups matching point pairs, eliminates misaligned pairs based on road topological connectivity, and establishes a mapping relationship between road lines and vector lines. Ultimately, through road line optimization and vector line optimization methods, we accomplish the automatic updating of vector data for mountainous roadways. Through a comprehensive analysis of over 2600 vector data, our method demonstrates substantial advantages in terms of both F1 and IoU compared to alternative methods.

A novel transformer‐based graph generation model for vectorized road design

AbstractRoad network design, as an important part of landscape modeling, shows a great significance in automatic driving, video game development, and disaster simulation. To date, this task remains labor‐intensive, tedious and time‐consuming. Many improved techniques have been proposed during the last two decades. Nevertheless, most of the state‐of‐the‐art methods still encounter problems of intuitiveness, usefulness and/or interactivity. As a rapid deviation from the conventional road design, this paper advocates an improved road modeling framework for automatic and interactive road production driven by geographical maps (including elevation, water, vegetation maps). Our method integrates the capability of flexible image generation models with powerful transformer architecture to afford a vectorized road network. We firstly construct a dataset that includes road graphs, density map and their corresponding geographical maps. Secondly, we develop a density map generation network based on image translation model with an attention mechanism to predict a road density map. The usage of density map facilitates faster convergence and better performance, which also serves as the input for road graph generation. Thirdly, we employ the transformer architecture to evolve density maps to road graphs. Our comprehensive experimental results have verified the efficiency, robustness and applicability of our newly‐proposed framework for road design.

Patched Line Segment Learning for Vector Road Mapping

This paper presents a novel approach to computing vector road maps from satellite remotely sensed images, building upon a well-defined Patched Line Segment (PaLiS) representation for road graphs that holds geometric significance. Unlike prevailing methods that derive road vector representations from satellite images using binary masks or keypoints, our method employs line segments. These segments not only convey road locations but also capture their orientations, making them a robust choice for representation. More precisely, given an input image, we divide it into non-overlapping patches and predict a suitable line segment within each patch. This strategy enables us to capture spatial and structural cues from these patch-based line segments, simplifying the process of constructing the road network graph without the necessity of additional neural networks for connectivity. In our experiments, we demonstrate how an effective representation of a road graph significantly enhances the performance of vector road mapping on established benchmarks, without requiring extensive modifications to the neural network architecture. Furthermore, our method achieves state-of-the-art performance with just 6 GPU hours of training, leading to a substantial 32-fold reduction in training costs in terms of GPU hours.

A traffic state prediction method based on spatial–temporal data mining of floating car data by using autoformer architecture

AbstractFloating car data (FCD), characterized by wide spatiotemporal coverage, low collection cost, and immunity to adverse weather conditions, are one of the key approaches for intelligent transportation systems to obtain real‐time urban road network traffic information. The research aims to utilize GPS data from taxis in Shanghai and vector geographic information data of the road network, with urban expressways as the research focus. Based on the different driving characteristics of expressways and the vehicles on the ramps below, a clustering analysis is employed to determine all floating vehicles traveling on the target road. Furthermore, an adaptive buffer zone consistent with the road orientation is established based on road vector geographic data. This allows for the extraction of FCD within segmented areas, and the average vehicle speed for that road segment is obtained through weighted calculations. This method fully exploits the natural characteristics of taxis in urban areas with a wide spatiotemporal distribution. The data effectiveness and coverage reach 90.2% and 85.7%, respectively, significantly surpassing the traditional grid‐based extraction method for FCD. Additionally, to capture the long‐term spatiotemporal dependencies of road network traffic states, a spatial–temporal autoformer (STAF) network based on spatial–temporal sequence autocorrelation is employed for traffic state prediction. The results indicate that the STAF method demonstrates good performance in medium‐ and long‐term prediction. We believe that the proposed FCD mining method in this paper provides a new approach for efficiently extracting large‐scale road network traffic states and conducting medium‐ to long‐term predictions.

Tiger Habitat Quality Modelling in Malaysia with Sentinel-2 and InVEST

Deforestation is a threat to habitat quality and biodiversity. In intact forests, even small levels of deforestation can have profound consequences for vertebrate biodiversity. The risk hotspots are Borneo, the Central Amazon, and the Congo Basin. Earth observation (EO) now provides regular, high-resolution satellite images from the Copernicus Sentinel missions and other platforms. To assess the impact of forest conversion and forest loss on biodiversity and habitat quality, forest loss in a tiger conservation landscape in Malaysia is analysed using Sentinel-2 imagery and the InVEST habitat quality model. Forest losses are identified from satellites using the random forest classification and validated with PlanetScope imagery at 3–5 m resolution for a test area. Two scenarios are simulated using InVEST, one with and one without the forest loss maps. The outputs of the InVEST model are maps of tiger habitat quality and habitat degradation in northeast Peninsular Malaysia. In addition to forest loss, OpenStreetMap road vectors and the GLC2000 land-cover map are used to model habitat sensitivity to threats from roads, railways, water bodies, and urban areas. The landscape biodiversity score simulation results fall sharply from ~0.8 to ~0.2 for tree-covered land cover when forest loss is included in the habitat quality model. InVEST makes a reasonable assumption that species richness is higher in pristine tropical forests than in agricultural landscapes. The landscape biodiversity score is used to compare habitat quality between administrative areas. The coupled EO/InVEST modelling framework presented here can support decision makers in reaching the targets of the Kunming-Montreal Global Biodiversity Framework. Forest loss information is essential for the quantification of habitat quality and biodiversity in tropical forests. Next generation ecosystem service models should be co-developed alongside EO products to ensure interoperability.

IS-RoadDet: Road Vector Graph Detection with Intersections and Road Segments from High Resolution Remote Sensing Imagery

IS-RoadDet: Road Vector Graph Detection with Intersections and Road Segments from High Resolution Remote Sensing Imagery

A novel framework for road vectorization and classification from historical maps based on deep learning and symbol painting

Road networks in the past are imperative for understanding evolution of transportation infrastructure, urban sprawl, and route planning, etc. Various approaches have been developed for road extraction from historical maps, among which deep learning techniques stand out as the most effective ones. However, little attention has been paid to investigating road vectorization and classification from historical maps. Moreover, road classification via machine learning methods usually requires large amounts of dedicated training data. To address these issues, this paper proposes a novel and comprehensive framework for road vectorization and classification on the basis of road segmentation from historical maps. First, deep learning is used to get pixel-wise raster road segmentation results, which are further skeletonized using morphological operations. Then, considering that each road class is represented with a certain symbol, a painting function is defined for each class able to paint the corresponding symbol. These painting functions are then used to draw road segments along the skeletons. Since the start and end points in each painting function are used to vectorise the segment, this method achieves vectorization and classification at the same time. Our method is validated on four Siegfried map sheets in Switzerland, and evaluated via both visual and quantitative assessments. The results indicate that the method is capable of classifying roads accurately. In particular, two evaluation metrics completeness and correctness achieve 90.69% and 72.71% respectively for road class 2 which accounts for the highest portion in the map. Moreover, the results of this method avoid the saw-toothed issue of vectorised road lines. This research is beneficial for creating complete vector road network datasets with class information to support decision-making in urban planning and transportation.

A Novel Vectorized Curved Road Representation Based Aerial Guided Unmanned Vehicle Trajectory Planning.

Unmanned vehicles frequently encounter the challenge of navigating through complex mountainous terrains, which are characterized by numerous unknown continuous curves. Drones, with their wide field of view and ability to vertically displace, offer a potential solution to compensate for the limited field of view of ground vehicles. However, the conventional approach of path extraction solely provides pixel-level positional information. Consequently, when drones guide ground unmanned vehicles using visual cues, the road fitting accuracy is compromised, resulting in reduced speed. Addressing these limitations with existing methods has proven to be a formidable task. In this study, we propose an innovative approach for guiding the visual movement of unmanned ground vehicles using an air-ground collaborative vectorized curved road representation and trajectory planning method. Our method offers several advantages over traditional road fitting techniques. Firstly, it incorporates a road star points ordering method based on the K-Means clustering algorithm, which simplifies the complex process of road fitting. Additionally, we introduce a road vectorization model based on the piecewise GA-Bézier algorithm, enabling the identification of the optimal frame from the initial frame to the current frame in the video stream. This significantly improves the road fitting effect (EV) and reduces the model running time (T-model). Furthermore, we employ smooth trajectory planning along the "route-plane" to maximize speed at turning points, thereby minimizing travel time (T-travel). To validate the efficiency and accuracy of our proposed method, we conducted extensive simulation experiments and performed actual comparison experiments. The results demonstrate the superior performance of our approach in terms of both efficiency and accuracy.

Vector Road Map Updating from High-Resolution Remote-Sensing Images with the Guidance of Road Intersection Change Detection and Directed Road Tracing

Updating vector road maps from current remote-sensing images provides fundamental data for applications, such as smart transportation and autonomous driving. Updating historical road vector maps involves verifying unchanged roads, extracting newly built roads, and removing disappeared roads. Prior work extracted roads from a current remote-sensing image to build a new road vector map, yielding inaccurate results and redundant processing procedures. In this paper, we argue that changes in roads are closely related to changes in road intersections. Hence, a novel changed road-intersection-guided vector road map updating framework (VecRoadUpd) is proposed to update road vector maps with high efficiency and accuracy. Road-intersection changes include the detection of newly built or disappeared road junctions and the discovery of road branch changes at each road junction. A CNN-based intersection-detection network (CINet) is adopted to extract road intersections from a current image and an old road vector map to discover newly built or disappeared road junctions. A road branch detection network (RoadBranchNet) is used to detect the direction of road branches for each road junction to find road branch changes. Based on the discovery of direction-changed road branches, the VecRoadUpd framework extracts newly built roads and removes disappeared roads through directed road tracing, thus, updating the whole road vector map. Extensive experiments conducted on the public MUNO21 dataset demonstrate that the proposed VecRoadUpd framework exceeds the comparative methods by 11.01% in pixel-level Qual-improvement and 13.85% in graph-level F1-score.

Research on Vector Road Data Matching Method Based on Deep Learning

Most of the existing vector data matching methods use traditional feature geometry attribute features to match, however, many of the similarity indicators are not suitable for cross-scale data, resulting in less accuracy in identifying objects. In order to solve this problem effectively, a deep learning model for vector road data matching is proposed based on siamese neural network and VGG16 convolutional neural network, and matching experiments are carried out. Experimental results show that the proposed vector road data matching model can achieve an accuracy of more than 90% under certain data support and threshold conditions.

Research on Urban Road Network Extraction Based on Web Map API Hierarchical Rasterization and Improved Thinning Algorithm

Most existing research on the vector road network is based on GPS trajectory travel information extraction, and urban GPS trajectory data are large and difficult to obtain. Based on this, this study proposes a road network extraction method based on network map API and designs a vector road network based on an improved image-processing algorithm using trajectory data. Firstly, a large number of trajectory data are processed by hierarchical rasterization. The trajectory points of the regional OD matrix are obtained by using the map API interface to generate the trajectory. Then, the image expansion processing is performed on the road network raster image to complete the information loss problem. The improved Zhang–Suen refinement algorithm is used to refine the idea to obtain the road center line, and the vector road network in the study area is obtained. Finally, taking the Harbin City of Heilongjiang Province as an example, compared with the road network of the network map, it has been demonstrated that using this technology may improve the traveler experience and the sustainability of urban traffic flow while reducing the number of manual procedures required, performing online incremental rapid change detection, and updating the present road network at a cheaper cost.

On Physical Urban Boundaries, Urban Sprawl, and Compactness Measurement: A Case Study of the Wen-Tai Region, China

China’s rapid urbanization has been accompanied by serious urban sprawl. Instead of measuring the physical urban boundaries (PUBs), most of existing studies in China rely on yearbook statistics to describe the growth of urbanized area; therefore, the understanding of the actual form and quantity of urban sprawl are restrained. As the statistical unit is generally at or above the county level, these studies tend to omit the lower-level “larger towns”. This paper discusses the measurement of urban sprawl and compactness using multi-source data on the GIS platform through the case study of the Wen-Tai region in China. GlobeLand30 remote sensing image data, vector road network data, NPP/VIIRS nighttime light data, and points of interest (POIs) data are adopted. The new method enhances the identification of built-up areas in larger towns. Besides, the 2020s’ PUBs of this region, data for 2010 and 2000 are retraced to assess the urban expansion rate, and two approaches are used to discuss the urban growth pattern. Additionally, a compactness model is constructed from four dimensions, i.e., the compactness of external contour, accessibility of road network, land-use intensity, and functional diversity, by which a high-resolution visual analysis tool is created for the provincial government to monitor urban sprawl.

Studying a matching method combining distance proximity and buffer constraints

Abstract In order to achieve fast matching of global navigation satellite system (GNSS) track data and vector road network data, a rough-to-fine matching method combining distance proximity and buffer constraints is studied. The GNSS track data is preprocessed: A rasterisation method is used to eliminate the low-quality track data and obtain the binary graph; then, the road centrelines are obtained by a skeleton extraction method; the thinning data is converted into line segments or closed lines starting from nodes by the boundary tracking method, which is stored in the vector form. Finally, the boundary tracking method is used to convert the thinning data into closed lines and store them in the vector form. A two-layer rough-to-fine matching strategy is applied: First, the changing state of the road arc is determined according to the rough matching strategy of road segment distance. Next, according to the buffer constraint criterion, a rectangular or circular buffer is selected to achieve precise matching of road network data. The results show that the proposed method has good matching effect, which is especially suitable for working with more GNSS track data.

Vector data partition correction method supported by deep learning

ABSTRACT Road vector data are an important part of geographic information databases and play a leading role in social and economic development. Among them, the accuracy and current situation of vector data are the key values of their application. In recent years, with the development of remote sensing technology, remote sensing images contain increasingly more road information, which provides a large number of available features for the establishment of vector correction models. Therefore, the vector-to-image vector correction method based on remote sensing images has become an important means to ensure the accuracy and current situation of vector data. However, in actual scenes, there are great differences in the design structure and design index between urban roads and rural roads, which is reflected in the serious spatial heterogeneity between them in the remote sensing images. Therefore, when the existing methods use remote sensing images for vector correction, the universality is limited by the change of regional scenes, so it is difficult to realize the simultaneous correction of urban road and rural road vectors, with poor applicability and low compatibility. To solve this problem, this paper proposes a vector data partition correction method supported by deep learning. First, the U2-net model and line segment sequence method are used to generate the feature set. Second, according to the characteristics of the poor quality of urban road extraction results, the regular geometric form of vector data and the poor structural information of road images, the methods of vector line decomposition, subvector line correction, subvector line centralization and vector line synthesis are proposed to correct the vector lines of urban roads. Finally, according to the characteristics of high-quality rural road extraction, the irregular geometric form of vector data and the strong information of road image structures, this paper proposes road edge extraction, a road centreline reasoning model and a vector line connection method to complete the correction of rural vector road data. The quantitative analysis of the experimental results shows that compared with other methods, the urban road vector correction method in this paper is much better than the comparison methods and has achieved better correction results in rural areas. This method has the advantages of compatibility and better accuracy for roads in different areas.

Geospatial Approach to Solving Traffic Congestion Problems In Mushin Local Government Area of Lagos State

The variation in road conditions in relation to traffic congestion overtime enhances the need for proper development of a geospatial platform by comprehensively building and relating the attributes information of the entire road network with respect to various road segments. This is better executed through the concept of dynamic road segmentation. The aim of this study is to carry out geospatial data structure of dynamic road segmentation in Mushin local government area of Lagos State showing the road segment (spatial data) and its topological information.The selected roads within Mushin Local Government were extracted from a street map of Mushin Local Government Area gotten from the office of the Surveyor General of Lagos state. The selected roads were digitized to obtain the spatial information required for the road segmentation. The road vectorization or digitization was carried out in ArcGIS 10.2 software. It was created as a separate shapefile and added as a layer within the GIS environment.The information acquired from the field survey describing the road network as well as the segments i.e. the attribute information, formed a database developed in a GIS platform using ArcGIS 10.2 software. This relational database contains information such as name of road, segment ID, nature of roads, nature of segments, presence of pot holes, presence of crime scenes, presence of congestion, causes of congestion, local inhabitants of segment, length of segments, adjoining land use (to the left and right), mode of transportation along the segment, the traffic condition of the roads and the number of lanes of the road.These large amount of attribute information for all the segment were geospatially referenced with respect to the positions of all the segments and this is the basis of every GIS operation; providing an environment for relating the spatial locations of features with the attribute information describing the spatial feature (which in this case the roads and the road segments represents the spatial feature)Spatial query by attribute as well as location can easily be carried out to select road segments of interest based on predefined criteria using a specific Structural Query Language (SQL) within the GIS environment.Dynamic road segmentation provides a platform for planning and controlling changes on road network. Since the roads are broken down into segments, temporal changes of the road structure as well as the traffic conditions of the roads can be analysed and suitable decisions as well as policies for proper road maintenance can easily be taken and the policies implemented with respect to the segmentation of the roads. It also aids the security organization implementing platform within a GIS environment to monitor the crime scenes with respect to affected segment and would aid in the deployment of security agents at the appropriate locations in case of emergency.In the control of traffic, dynamic road segmentation aid in deciding the alternative conditions to follow since the entire road network have being spatially segmented and the conditions of all the segments built into the GIS database.

Exploring Built Environment Influence on Taxi Vacant Time in Megacities: A Case Study of Chongqing, China

Driven by the emission peak and carbon neutrality targets, traditional cruising taxis are also under pressure to reduce emissions, and reducing taxi vacant time is an effective way to reduce emissions. This paper aims to examine the influence mechanism of built environment characteristics on vacant time of taxi trips from the supply-side perspective. To do so, the study uses a one week of taxi trajectory data of 2019 provided by a taxi company in Chongqing, a megacity in China. We take the divided square grid cell as the study units, calculate the taxi vacant time of 168 (7 × 24) hourly slots for each grid, and then calculate the built environment indicators of the grids with Point of Interest (POI) and vector road network data. A generalized additive model (GAM) based on panel data is constructed to investigate the influence mechanism of built environment on taxi vacant time. The results show that the time-varying trend of taxi vacant time is opposite to the time-varying trend of taxi trips and is similar to taxi trips in spatial distribution with significant spatial dependence. Taxi vacant time is negatively correlated with taxi trips. POI mixture, the number of enterprises, governmental agencies, and shopping services are negatively correlated with vacant time, while the number of sports and leisure services, hotel services, healthcare services, financial and insurance services, and living services are positively correlated with vacant time. An interesting finding is that metro stations significantly increase the vacant taxi time in the region. Average traffic flow speed, road density, the number of bus stops, catering services, residence communities, and motor vehicle services show strong nonlinear relationships with taxi vacant time, and their influence effects are alternately positive and negative. The study provides useful insights for understanding mechanisms of the role of built environment on taxi vacant time and has important implications for driver searching strategy improvement, taxi management measures development, spatial and temporal scheduling of taxi capacity resources, and urban transportation facilities layout planning.

RoadVecNet: a new approach for simultaneous road network segmentation and vectorization from aerial and google earth imagery in a complex urban set-up

ABSTRACT In this study, we present a new automatic deep learning-based network named Road Vectorization Network (RoadVecNet), which comprises interlinked UNet networks to simultaneously perform road segmentation and road vectorization. Particularly, RoadVecNet contains two UNet networks. The first network with powerful representation capability can obtain more coherent and satisfactory road segmentation maps even under a complex urban set-up. The second network is linked to the first network to vectorize road networks by utilizing all of the previously generated feature maps. We utilize a loss function called focal loss weighted by median frequency balancing (MFB_FL) to focus on the hard samples, fix the training data imbalance problem, and improve the road extraction and vectorization performance. A new module named dense dilated spatial pyramid pooling, which combines the benefit of cascaded modules with atrous convolution and atrous spatial pyramid pooling, is designed to produce more scale features over a broader range. Two types of high-resolution remote sensing datasets, namely, aerial and Google Earth imagery, were used for road segmentation and road vectorization tasks. Classification results indicate that the RoadVecNet outperforms the state-of-the-art deep learning-based networks with 92.51% and 93.40% F1 score for road surface segmentation and 89.24% and 92.41% F1 score for road vectorization from the aerial and Google Earth road datasets, respectively. In addition, the proposed method outperforms the other comparative methods in terms of qualitative results and produces high-resolution road segmentation and vectorization maps. As a conclusion, the presented method demonstrates that considering topological quality may result in improvement of the final road network, which is essential in various applications, such as GIS database updating.

Urban road mapping based on an end-to-end road vectorization mapping network framework

Reliable urban road vector maps are essential for urban analysis because the spatial distribution of road networks reflects urban development under the combined effects of nature and socio-economics. Diverse very high resolution (VHR) remote sensing images are now available, enabling explicit extraction of urban road vector maps over wide areas. Urban road vectorization mapping consists of two separate tasks: road extraction and road vectorization. The traditional methods focus on the road extraction task, and can obtain a good performance when using a pixel-based metric. However, the road vectorization methods are faced with the problem of road connectivity. In this work, to implement urban road vectorization mapping in a unified way, an end-to-end road vectorization mapping network (RVMNet) framework is proposed. The proposed RVMNet framework consists of a node proposal network (NPN) module and a node connectivity based road refinement module. In the NPN module, a fully convolutional network is adopted for the road node extraction. This improves the connectivity of the road mask by providing supervised information in the form of the road nodes, which are actually part of the road mask. The road mask is then converted into a road vector map by vectorization. In the node connectivity based road refinement module, road nodes are inserted into the road vector map to improve the connectivity. We compared RVMNet with the other state-of-the-art road detection methods on two public road datasets (SpaceNet 3 and DeepGlobe). The results of this comparison showed that combining road extraction and road vectorization into a unified framework is an efficient and accurate strategy for urban road vectorization mapping because it can propose road nodes that help to improve the road connectivity. Moreover, we constructed the novel UrbanRoadNet dataset, covering six cities (Beijing city center; Helsinki; Wuhan; Macao; the Wan Chai area of Hong Kong; and Shanghai). We then applied the RVMNet framework to the data from the six cities, obtaining an improvement in the vector-based average path length similarity (APLS) value of 4.1%. The spatial transfer assessments from both the qualitative and quantitative aspects corroborated the robust generalizability of the proposed method, and further verified the effectiveness of the proposed approach for large-scale road vectorization mapping at a very high resolution. It was also found that road vector spatial distribution is a useful way to reflect urban development.

Mapping large-scale and fine-grained urban functional zones from VHR images using a multi-scale semantic segmentation network and object based approach

Urban functional zones (UFZs) are essential for characterizing both urban spatial configurations and socio-economic properties and monitoring urbanization process, thus UFZs are fundamental to urban planning, management and renewal. Although many efforts in remote sensing field have been made to map UFZs, large-scale and fine-grained UFZ maps required by a broad range of urban applications are still unavailable. Existing methods generally rely on pre-determined mapping units, such as image tiles and road blocks, which significantly limit the mapping quality and the automation degree of mapping UFZs. Given that UFZs are composed of diverse geographic objects, this study proposes a novel object-based UFZ mapping method using very-high-resolution (VHR) remote sensing images. First, a multi-scale semantic segmentation network that achieves pixel-wised predictions is proposed to predict urban-functions for geographic objects by capturing multi-scale contextual information. Afterwards, a conditional random field (CRF) framework is designed to regroup objects into UFZs to produce the final UFZ map, wherein road vectors are incorporated to restrict the procedure. The presented object-as-analysis-unit scheme conquers the drawbacks of mapping-unit pre-determination and the semantic segmentation model provides accurate function information for objects, thus they can be applied for producing large-scale and fine-grained UFZ maps. In the experiment, the proposed method is evaluated by producing UFZ maps for Beijing and Shanghai, China, and competitive results with overall accuracy of 91.6% and 89.1% are achieved, respectively. Finally, the generated UFZ maps are utilized to analyze the urban-function structures of the two cities. The proposed method can be regarded as a significant development that appears to be promising and practical for mapping UFZ maps for real-world urban applications.