Urban Mapping Research Articles

An Assessment of Cultural Places of Interest in Tshwane to Inform a Tourism Shift From Colonial to Indigenous Towards Decolonisation

This paper continues from the premise that the built form of the South African post-apartheid city continues to perpetuate the paradigm that only the colonial culture is available for tourist consumption. South Africa is a culturally diverse country that provides tourism to an international tourist market that seeks diverse cultural experiences. However, with most African cultural artefacts found encased in colonial architecture museums, questions arise regarding the authenticity of the cultural experience and the appropriate cultural representation in built form. This paper provides recommendations towards the transformation of the post-apartheid city following an assessment of the current tourist attractions that host cultural artefacts and serve as tourist attractions. Most of this data exists, and this paper investigates, collates and assesses the data based on these criteria: colonial and indigenous, restricted, and accessible, and static or transformative. The collated data is presented in urban mapping and architectural illustrations. The study is focused on the City of Tshwane in South Africa. Selected case studies are presented. The findings indicate a substantial under-representation in the City of Tshwane of what this paper argues to be culturally appropriate tourist attractions. This paper argues that there is a market for the consumption of culture as an experience away from colonial towards indigenous.

A novel urban vegetation mapping approach for fire risk assessment: A Mediterranean case study

A novel urban vegetation mapping approach for fire risk assessment: A Mediterranean case study

Vehicle Trajectory Prediction via Urban Network Modeling.

Taxis are an important component of the transportation system, and empty taxis represent a significant waste of transportation resources. To alleviate the imbalance between supply and demand and relieve traffic congestion, real-time prediction of taxi trajectories is necessary. Most existing trajectory prediction studies focus on extracting time-series information but do not capture spatial information sufficiently. In this paper, we focus on the construction of an urban network and propose an urban topology-encoding spatiotemporal attention network (UTA) to address destination prediction problems. Firstly, this model discretizes the production and attraction units of transportation, combining them with key nodes in the road network to form an urban topological network. Secondly, GPS records are matched to the urban topological map to construct a topological trajectory, which significantly improves trajectory consistency and endpoint certainty, helping to model destination prediction problems. Thirdly, semantic information concerning surrounding space is attached to effectively mine the spatial dependencies of trajectories. Finally, after the topological encoding of city space and trajectories, this algorithm proposes a topological graph neural network to model the attention calculation with the trajectory context, comprehensively considering the spatiotemporal characteristics of the trajectories and improving prediction accuracy. We solve the prediction problems with the UTA model and also compare it with some other classical models, such as the HMM, RNN, LSTM, and transformer. The results suggest that all the models work well in combination with the proposed urban model (with a rough increase of 2%), while the UTA model is less affected by data sparsity.

Examining effect of super-resolution on AVIRIS-NG data: A precursor to generation of large-scale urban material and natural cover maps

The two-phase Airborne Visible InfraRed Imaging Spectrometer – Next Generation (AVIRIS-NG) campaign over India has generated datasets of numerous natural and artificial surroundings with high signal-to-noise ratios and higher spatial and spectral consistencies. However, the first phase AVIRIS-NG datasets of urban areas have 8.1 m spatial resolution, which is insufficient for producing level-4 urban material and natural cover maps required to model Earth’s processes at the micro-scale or constantly monitor urban processes. The current work addresses this drawback by demonstrating a part of the proposed proof-of-concept approach on the AVIRIS-NG dataset of Ahmedabad, India, wherein super-resolution (SR) is proposed to enhance the hyperspectral image’s spatial resolution and preserve its rich spectral content at the targeted spatial scale before utilizing it for producing detailed urban land cover maps. A step-wise account of the SR process has been provided herein to aid the reader in reproducing these experiments. Super-resolved products generated using four SR algorithms have been compared and validated using a composite framework based on processing time, qualitative visual investigation of selected image patches and spectral profiles, re-interpretation of Wald’s reduced resolution protocol and full-reference image quality metrics. Results suggest the AVIRIS-NG dataset’s successful spectral and spatial fidelity at a larger spatial scale, albeit varying in each super-resolved output. Sparse regression and natural prior (SRP), and anchored neighbourhood regression (ANR) produce the best and poorest super-resolved outputs among the tested SR algorithms. Although gradient profile prior (GPP) generated super-resolved product is the second best, it is also the most computationally expensive algorithm. An additional yardstick in the form of a spectral similarity analysis of randomly selected unknown spectral signatures on super-resolved output against a reference spectral library has been proposed to ascertain spectral content preservation further before utilizing the super-resolved product for urban information extraction. This exercise’s findings reveal the presence of the target urban land cover feature at the unknown pixels with matching scores of greater than 0.5 in both best and worst super-resolved products, thereby advocating the extension of the proposed proof-of-concept to recently launched spaceborne hyperspectral missions’ datasets.

GIS Based Urban Noise Pollution Analysis and Mapping: The Case of Antalya, Turkey

In this study, it is aimed to determine the urban noise level, to classify it according to the provisions of the relevant regulation, to map it using geographic information systems (GIS) and to develop physical and administrative suggestions for eliminating the noise above the threshold value. In this context, the city center of Antalya was chosen as the study area. This area was digitized, and its maps were produced with the help of remote sensing and geographic information systems. Then, the noise levels of the region were measured using a decibel meter, provided that it was evenly distributed throughout the study area, all measurement points were coordinated, transferred to the map and a database was created. Interpolation was applied based on the noise threshold values in the "Evaluation and Management of Environmental Noise" regulation in the national legislation, noise levels were classified and noise maps were produced. In accordance with the relevant regulation, land survey was carried out again for the regions exceeding the noise limit values, and the sources of noise were identified. In line with the results, physical and administrative solutions have been developed for the elimination of noise in the region.

A labor-free index-guided semantic segmentation approach for urban vegetation mapping from high-resolution true color imagery

ABSTRACT Accurate and timely information on urban vegetation (UV) can be used as an important indicator to estimate the health of cities. Due to the low cost of RGB cameras, true color imagery (TCI) has been widely used for high spatial resolution UV mapping. However, the current index-based and classifier-based UV mapping approaches face problems of the poor ability to accurately distinguish UV and the high reliance on massive annotated samples, respectively. To address this issue, an index-guided semantic segmentation (IGSS) framework is proposed in this paper. Firstly, a novel cross-scale vegetation index (CSVI) is calculated by the combination of TCI and Sentinel-2 images, and the index value can be used to provide an initial UV map. Secondly, reliable UV and non-UV samples are automatically generated for training the semantic segmentation model, and then the refined UV map can be produced. The experimental results show that the proposed CSVI outperformed the existing five RGB vegetation indices in highlighting UV cover and suppressing complex backgrounds, and the proposed IGSS workflow achieved satisfactory results with an OA of 87.72% ∼ 88.16% and an F1 score of 87.73% ∼ 88.37%, which is comparable with the fully-supervised method.

Causes and Effects of Urban Sprawl: A Case Study of Jammu City

Abstract: Jammu city has faced unprecedented growth of population. Before fifties Jammu was small town but after 1971 it starts growing both in size and shape. Due to migration from valley and rapid industrialization (between 1989 to 1994) the population of Jammu grew up to 7, 27,940 representing 225.90% increases over 1981 population. The focus of this paper is on the cause and effects of urban sprawl through landsat images. Remote sensing and GIS techniques have been used to analyze the urban sprawl mapping and detect changes of urban Sprawl of Jammu city through different year. Satellite data have been extracted from different sites which are useful in mapping and quantifying the extent of urban area of Jammu city in different time periods. Urban Sprawl is one of the burning issues in Jammu that may boost up at a vigorous rate after 1972 the land use under the agriculture continually declines at a faster rate which presents the threat to the food security for human life. At present Jammu city has sprawling towards its all direction along the main transport routes. In this study Geographical information system (GIS) and different satellites images have been used to extract the buildup area of different year (from 1972 to 2017) which is useful for urban planning.

Multi-featured multi-scale combination of high-resolution remote sensing images for building extraction

Abstract As one of the important feature categories in urban geographic data, buildings are the key thematic elements to be represented in large-scale urban mapping with the high speed of urban digital construction. The identification and extraction of buildings are of great significance for feature extraction, feature matching, image interpretation and mapping. However, the great variability of building size, shape, color, orientation, etc., in remote sensing images poses a great challenge to building detection. To this end, this paper proposes an algorithm based on multi-feature multi-scale fusion for the automatic extraction of buildings in remote sensing images are represented in the form of roofs. It is difficult to represent all buildings with a single feature because of the different colors, textures and shapes of building roofs. Effective features to describe buildings are proposed, including edge density and edge distribution, brightness contrast, color contrast and other features to describe building edge brightness. We propose effective features to describe buildings, including edge density and edge distribution, luminance contrast, color contrast and other underlying features to describe the edges, luminance and color of buildings, and adding special structural features such as main direction orthogonality and target integrity and symmetry to describe buildings by multiple features together. Moreover, the K-value nearest neighbor classification algorithm is used to train a series of samples, and the weights of each feature in the multi-feature model are obtained through iterative learning to obtain the multi-feature linear model and calculate the visual saliency of buildings in the sliding window; finally, the proposed algorithm has experimented with several groups of high-resolution remote sensing images respectively, and the multi-scale multi-feature fusion model algorithm is used as the Erkoff random field model to compare the algorithm. The results of this paper show that the proposed multiscale multi-feature fusion model algorithm improves by 10.82% for building classification accuracy extraction and 13.96% for feature selection extraction accuracy, and finally, the comparison from the shape optimization effect figure concludes that the multiscale multi-feature fusion model can achieve better extraction accuracy and practical effect for buildings in remote sensing images, which has certain practicality and It has certain practicality and superiority. It promotes the in-depth application of multi-feature multi-scale combined high-resolution remote sensing image-building extraction in geographic states, road traffic and other industries.t

A graph-based framework to integrate semantic object/land-use relationships for urban land-use mapping with case studies of Chinese cities

Urban land-use types, such as residential and administration, can be inferred through semantic objects and their relationships. Point of interest (POI) data can serve as the semantic objects for urban land-use mapping. However, the previous POI-based approaches have rarely considered the relationships between the semantic objects in the urban land-use mapping, and three main challenges remain: 1) the lack of paired semantic object/land-use samples; 2) the lack of a unified model for semantic objects and the relationships between sematic objects and urban land use; and 3) the difficulty of automatically learning semantic object/land-use mapping relationships. In this paper, to address these issues, a graph-based urban land-use mapping framework integrating semantic object/land-use relationships (GOLR) is proposed. Based on open-source area of interest (AOI) and POI data, an urban object/land-use (UOLU) dataset covering 34 cities in China was built. To model the spatial and mapping relationships, the semantic objects and their relationships are used to jointly build an urban land-use graph. The mapping from semantic objects to urban land use can then be learned by the urban land-use graph isomorphic network (ULGIN) model. Finally, the GOLR framework was applied to obtain accurate land-use mapping results for multiple Chinese cities.

Data fusion uncertainty-enabled methods to map street-scale hourly NO2 in Barcelona: a case study with CALIOPE-Urban v1.0

Abstract. Comprehensive monitoring of NO2 exceedances is imperative for protecting human health, especially in urban areas with traffic. However, an accurate spatial characterization of the exceedances is challenging due to the typically low density of air quality monitoring stations and the inherent uncertainties in urban air quality models. We study how observational data from different sources and timescales can be combined with a dispersion air quality model to obtain bias-corrected NO2 hourly maps at the street scale. We present a kriging-based data fusion workflow that merges dispersion model output with continuous hourly observations and uses a machine-learning-based land use regression (LUR) model constrained with past short intensive passive dosimeter campaign measurements. While the hourly observations allow the bias adjustment of the temporal variability in the dispersion model, the microscale LUR model adds information on the NO2 spatial patterns. Our method includes an uncertainty calculation based on the estimated error variance of the universal kriging technique, which is subsequently used to produce urban maps of probability of exceeding the 200 µg m−3 hourly and the 40 µg m−3 annual NO2 average limits. We assess the statistical performance of this approach in the city of Barcelona for the year 2019. Our results show that simply merging the monitoring stations with the model output already significantly increases the correlation coefficient (r) by +29 % and decreases the root mean square error (RMSE) by −32 %. When adding the time-invariant microscale LUR model in the data fusion workflow, the improvement is even more remarkable, with +46 % and −48 % for the r and RMSE, respectively. Our work highlights the usefulness of high-resolution spatial information in data fusion methods to better estimate exceedances at the street scale.

Fine-scale mapping of urban malaria exposure under data scarcity: an approach centred on vector ecology

BackgroundAlthough malaria transmission has experienced an overall decline in sub-Saharan Africa, urban malaria is now considered an emerging health issue due to rapid and uncontrolled urbanization and the adaptation of vectors to urban environments. Fine-scale hazard and exposure maps are required to support evidence-based policies and targeted interventions, but data-driven predictive spatial modelling is hindered by gaps in epidemiological and entomological data. A knowledge-based geospatial framework is proposed for mapping the heterogeneity of urban malaria hazard and exposure under data scarcity. It builds on proven geospatial methods, implements open-source algorithms, and relies heavily on vector ecology knowledge and the involvement of local experts.MethodsA workflow for producing fine-scale maps was systematized, and most processing steps were automated. The method was evaluated through its application to the metropolitan area of Dakar, Senegal, where urban transmission has long been confirmed. Urban malaria exposure was defined as the contact risk between adult Anopheles vectors (the hazard) and urban population and accounted for socioeconomic vulnerability by including the dimension of urban deprivation that is reflected in the morphology of the built-up fabric. Larval habitat suitability was mapped through a deductive geospatial approach involving the participation of experts with a strong background in vector ecology and validated with existing geolocated entomological data. Adult vector habitat suitability was derived through a similar process, based on dispersal from suitable breeding site locations. The resulting hazard map was combined with a population density map to generate a gridded urban malaria exposure map at a spatial resolution of 100 m.ResultsThe identification of key criteria influencing vector habitat suitability, their translation into geospatial layers, and the assessment of their relative importance are major outcomes of the study that can serve as a basis for replication in other sub-Saharan African cities. Quantitative validation of the larval habitat suitability map demonstrates the reliable performance of the deductive approach, and the added value of including local vector ecology experts in the process. The patterns displayed in the hazard and exposure maps reflect the high degree of heterogeneity that exists throughout the city of Dakar and its suburbs, due not only to the influence of environmental factors, but also to urban deprivation.ConclusionsThis study is an effort to bring geospatial research output closer to effective support tools for local stakeholders and decision makers. Its major contributions are the identification of a broad set of criteria related to vector ecology and the systematization of the workflow for producing fine-scale maps. In a context of epidemiological and entomological data scarcity, vector ecology knowledge is key for mapping urban malaria exposure. An application of the framework to Dakar showed its potential in this regard. Fine-grained heterogeneity was revealed by the output maps, and besides the influence of environmental factors, the strong links between urban malaria and deprivation were also highlighted.

A Framework on Fast Mapping of Urban Flood Based on a Multi-Objective Random Forest Model

Fast and accurate prediction of urban flood is of considerable practical importance to mitigate the effects of frequent flood disasters in advance. To improve urban flood prediction efficiency and accuracy, we proposed a framework for fast mapping of urban flood: a coupled model based on physical mechanisms was first constructed, a rainfall-inundation database was generated, and a hybrid flood mapping model was finally proposed using the multi-objective random forest (MORF) method. The results show that the coupled model had good reliability in modelling urban flood, and 48 rainfall-inundation scenarios were then specified. The proposed hybrid MORF model in the framework also demonstrated good performance in predicting inundated depth under the observed and scenario rainfall events. The spatial inundated depths predicted by the MORF model were close to those of the coupled model, with differences typically less than 0.1 m and an average correlation coefficient reaching 0.951. The MORF model, however, achieved a computational speed of 200 times faster than the coupled model. The overall prediction performance of the MORF model was also better than that of the k-nearest neighbor model. Our research provides a novel approach to rapid urban flood mapping and flood early warning.

Towards a detailed urban structural mapping using outcrop, geophysical and borehole data integration: Implications for sustainable development

Nowadays, geological conditions and geohazards assessment are essential requirements for sustainable urban planning and development. Hence, with lack of outcrop and near-surface geology, additional geohazards can be imposed. The present study is the first such challenge in one of highly deformed areas in Egypt and, accordingly, an integrated approach for detailed urban geological and structural mapping is given. Therefore, the surface and subsurface field measurements converge on the detailed geological conditions characterization including structural settings in a new urban city, as a case study. In this regard, the measured surface structural data are presented on a detailed geological map considering the surface rock units geometry and distributions. In order to unravel the near-surface geology, geophysical direct current resistivity (DCR) sounding survey is performed to represent subsurface layers distribution considering the surface geological mapping. To reduce the limitations of DCR sounding interpretation, conventional and non-conventional inversion schemes are applied with reference to available borehole data. Evidently, the sum of surface and near-surface mapped fault segments are plotted and analyzed to provide a detailed geometrical structural model. Because the fault rupture hazards and shale layer expansion issues are closely related to the urban sustainable development, a preliminary geohazards assessment is illustrated in the form of fault setbacks calculation and the shale free swelling evaluation. The present approach provides a holistic view of the urban geological mapping and introduces a means of recognizing preliminary potential geohazards problems and/or opportunities at an early stage in any proposed sustainable development, which can support site investigation designing strategies.

Satellite Image Processing Using Radiometric Resolution & Spectral Resolution

Abstract: This paper presents a detailed comparison of various image processing techniques for analysing satellite images. The satellite images are large in size, acquired from long distances and are affected by noise and other environmental conditions. Hence it is necessary to process them so that they can be used by the researchers for analysis. Spectral resolution basically is to measure changes in things that impact our environments like water quality or vegetation etc. Satellite images are widely used in many real time applications such as in agriculture land detection, navigation and in geographical information systems. In this paper, a review of spectral resolution requirements for urban mapping evaluated how spectral resolution of high-spatial resolution optical remote sensing data influences detailed mapping of urban land cover. A comprehensive regional spectral library and low altitude data from the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) were used to characterize the spectral properties of urban land cover.

URBAN MAP GENERATION IN ARTIST’S STYLE USING GENERATIVE ADVERSARIAL NETWORKS (GAN)

Artificial Intelligence is a field that is able to learn from existing data to synthesize new ones using deep learning methods. Using Artificial Neural Networks that process big datasets, complex tasks and challenges become easily resolved. As the zeitgeist suggests, it is possible to produce novel outcomes for future projections by applying various machine learning algorithms on the generated data sets. In that context, the focus of this research is exploring the reinterpretation of 21st century urban plans with familiar artist styles using different subtypes of deep-learning-based generative adversarial networks (GAN) algorithms. In order to explore the capabilities of urban map transformation with machine learning approaches, two different GAN algorithms which are cycleGAN and styleGAN have been applied on the two main data sets. First data set, the urban data set, contains 50 cities urban plans in .jpeg format collected according to the diversity of the urban morphologies. Whereas the second data set is composed of four well-known artist’s paintings, that belong to various artistic movements. As a result of training the same data sets with different GAN algorithms and epoch values were compared and evaluated. In this respect, the study not only investigates the reinterpretation of stylistic urban maps and shows the discoverability of new representation techniques, but also offers a comparison of the use of different image to image translation GAN algorithms.

Development of Mobile Information Collection System Based on IServer

Data is the "blood" of GIS. GIS focuses on data input, analysis and output, so data collection is particularly important. Traditional surveying and mapping has some shortcomings, such as backward equipment, low accuracy of data acquisition, low acquisition efficiency, and high cost of manpower and material resources. In the information collection of modern urban planning, traditional surveying and mapping has been difficult to meet the needs of data collection. At the same time, the emergence of various mobile intelligent devices and the increasingly mature development of related software and hardware make mobile terminal information acquisition technology gradually replace traditional surveying and mapping in field operations. This paper combines traditional surveying and mapping and modern information technology to develop the mobile terminal information acquisition system based on iServer. First, learn about the development of information collection technology in mobile GIS at home and abroad by consulting a large number of relevant literature, and define the value and significance of developing mobile terminal information collection system. Then, analyze the shortcomings of traditional surveying and mapping in field survey and the problems existing in mobile GIS collection information at present, combine available resource tools to form a project demand analysis, and conceive a project outline design according to the demand analysis, Finally, the information collection system based on iServer and Android is designed in detail, including software functional framework, database design and server side functional design. Theoretical analysis and practical results show that the system can effectively improve the accuracy and speed of data acquisition while reducing the cost of data acquisition.

Flood Susceptibility Mapping Using Image-Based 2D-CNN Deep Learning: Overview and Case Study Application Using Multiparametric Spatial Data in Data-Scarce Urban Environments

This study presents results for urban flood susceptibility mapping (FSM) using image-based 2D-convolutional neural networks (2D-CNN). The model input multiparametric spatial data comprised of land-useland-cover (LULC), digital elevation model (DEM), and the topographic and hydrologic conditioning derivatives, precipitation, and soil types. The implemented dropout regularization 2D-CNN with ReLU activation function, categorical cross-entropy loss function, and AdaGrad optimizer produced the case study area FSM with overall accuracy (OA) of 82.5%. The image-based 2D-CNN outperformed the multilayer perceptron (MLP) neural network by 18.4% in terms of overall accuracy and with corresponding lower MAE and higher F1-measures of 10.9% and 0.989, as compared to 25.6% and 0.877, respectively, for MLP-ANN results. The accuracy of the 2D-CNN that produced FSM map and the model efficiency were evaluated using area under the ROC curve (AUC) with respective success and prediction rates of 0.827 and 0.809. Using image-based 2D-CNN, 27% of the 247.7 km2 of the studied area was mapped with a high risk of flooding, with MLP-ANN overestimating the degree of high flood risk by 4.7%. Based on the gain ratio index analysis of the flood conditioning factors (FCFs), the most significant FCFs were LULC (18.5%), precipitation (14.9%), proximity to river (13.3%), and elevation (12.4%). Soil types contributed 8.6%, slope 9.1%, and the DEM-derived hodological conditioning indicators contributed 23.2%. The study results demonstrate that in urban areas with scarce hydrological monitoring networks, the use of image-based 2D-CNN with multiparametric spatial data can produce high-quality flood susceptibility maps for flood management in urban environments.

LANDSAT 8 LST Pan sharpening using novel principal component based downscaling model

The panchromatic (PAN) band is accessible at a much higher spatial resolution than the thermal infrared (TIR) band in remote sensing satellites. To mitigate this issue, current research proposes a fusion method based on principal components of PAN and TIR band pairs to enhance the spatial resolution of Land surface temperature (LST) images. In this research, LANDSAT-8 LST images of 100 m resolution have been downscaled to 15 m resolution by the Novel principal component (NPC) based downscaling model. The downscaled LST image's thermal values have been cross-verified by in situ thermal points collected by the thermal data logger and image quality indices. The results have shown the root mean square error (RMSE) of 0.51–0.92 for downscaled LST image from 100m to 15m resolution. The correlation between downscaled LST and ground truth points is between 0.54 and 0.83. The proposed NPC-based LST-PAN combination technique collects spatial details of the Earth's surface from PAN data as well as indulges thermal information of TIR image. The proposed fusion technique has upgraded the spatial goal and peculiarity of TIR images data set. The higher spatial resolution of thermal images can increase the level of precision in various remote sensing applications. These improved data sets can be used in a variety of applications, including urban heat island mapping, agricultural crop monitoring, forest fire detection, and hydrological modeling, to name a few. The increased spatial resolution provided by NPC fusion results in more accurate and consistent temperature readings, leading to improved decision-making and management in these and other applications.

Developing an intelligent cloud attention network to support global urban green spaces mapping

Urban green spaces (UGS) play an important role in understanding of urban ecosystems, climate, environment, and public health concerns. Satellite derived UGS maps provide an efficient and effective tool for urban studies and contribute to targets and indicators of the sustainable development goals, at the global level, set by the United Nations. However, clouds create a challenging issue in optical satellite image processing, leading to significant uncertainty in UGS mapping. In this study, we propose an automatic UGS mapping method by integrating satellite images with crowdsourced geospatial data while aiming to reduce the uncertainty caused by cloud contamination. The proposed method consists of three parts: (1) auxiliary data pre-processing module; (2) cloud attention intelligent network (CAI-net); and (3) non-cloud scenes classification module. The auxiliary data pre-processing module was used to convert crowdsourcing geospatial data into auxiliary maps. The CAI-net was proposed to retrieve detailed UGS classes within clouds from satellite image patches and auxiliary maps, while non-cloud scenes classification module was used to extract UGS from satellite image patches. The proposed method was applied to generate spatial continuous global UGS map products, considering the uncertainty caused by cloud contamination. The results show the proposed method yielded a high-quality global UGS map with average overall accuracy as high as 92.96% when satellite images had cloud coverage ranging from 0% to 50%. The geospatial AI, specifically CAI-net, can provide more accurate UGS mapping regardless of different geographical and climatic conditions of the study areas, which is especially significant for humid tropical and subtropical regions with frequent clouds and rains.

SemiRoadExNet: A semi-supervised network for road extraction from remote sensing imagery via adversarial learning

Road extraction from remote sensing imagery is a popular and frontier research focus, since road information plays an essential role in application fields, such as urban management, map updating and traffic planning. Deep learning based methods have shown their dominance on road extraction from remote sensing imagery. However, the performance of the existing road extraction methods relies heavily on a large amount of high-quality annotated training data, which is usually hard to obtain in practice. Current semi-supervised road extraction models can effectively reduce the dependency on the labeled data, nevertheless they cannot fully utilize the latent information of low-confidence pixels in pseudo-labels effectively. These pixels are usually a border between road and non-road area and of importance for the prediction accuracy of road extraction models. In order to address these issues, we proposed a novel semi-supervised road extraction network, named SemiRoadExNet. SemiRoadExNet is based on a Generative Adversarial Network (GAN), containing one generator with two discriminators. Firstly, both labeled and unlabeled images are put into the generator network for road extraction, and the outputs of the generator not only include road segmentation results but also the corresponding entropy maps. The entropy maps represent the confidence of prediction (road or non-road) for each pixel. Then, the two discriminators enforce the feature distributions keeping the consistency of road prediction maps and entropy maps between the labeled and unlabeled data. During the adversarial training, the generator is continuously regularized by exploiting the potential information from unlabeled data, thus the generalization capacity of the proposed model can be improved effectively. Compared to several state-of-the-art semi-supervised semantic segmentation methods, the proposed SemiRoadExNet achieves 0.96–5.38% IoU improvements on DeepGlobe Road Extraction, Massachusetts Roads and CHN6-CUG datasets respectively. The source code of d SemiRoadExNet is freely available at https://github.com/hchen118/SemiRoadExNet.