Search

Change detection (CD) using deep learning techniques is a prominent topic in the field of remote sensing (RS). However, the existing methods require large amounts of labeled samples for supervised learning, which is time-consuming and labor-intensive. To address this challenge, semi-supervised learning methods that utilize a limited number of labeled samples along with a large pool of unlabeled samples have emerged as a compelling solution. We propose a novel semi-supervised CD (SSCD) network that combines self-training and consistency regularization, namely STCRNet. During the self-training phase, STCRNet selects unlabeled samples with reliable pseudolabels based on their prediction stability across different training epochs and the consistency between class activation maps and prediction results within the model. Then, we apply data augmentation to the reliable samples and enforce consistency regularization on the augmented samples using the pseudolabels to enhance the network's robustness. Moreover, feature consistency regularization is applied to the remaining unlabeled samples with image perturbations, thereby broadening the feature space and improving the model's generalization performance. The experimental results on two widely used datasets demonstrate that STCRNet achieves state-of-the-art performance, especially with a significantly small amount (5%–10%) of labeled samples. STCRNet presents a promising solution for SSCD.

Convolutional neural networks have demonstrated remarkable capability in extracting deep semantic features from images, leading to significant advancements in various image processing tasks. This success has also opened up new possibilities for change detection (CD) in remote sensing applications. But unlike the conventional image recognition tasks, the performance of AI models in CD heavily relies on the method used to fuse the features from two different phases of the image. The existing deep-learning-based methods for CD typically fuse features of bitemporal images using difference or concatenation techniques. However, these approaches often fail tails to prioritize potential change areas adequately and neglect the rich contextual information essential for discerning subtle changes, potentially leading to slower convergence speed and reduced accuracy. To tackle this challenge, we propose a novel feature fusion approach called feature-difference attention-based feature fusion CD network. This method aims to enhance feature fusion by incorporating a feature-difference attention-based feature fusion module, enabling a more focused analysis of change areas. Additionally, a deep-supervised attention module is implemented to leverage the deep surveillance module for cascading refinement of change areas. Furthermore, an atrous spatial pyramid pooling fast is employed to efficiently acquire multiscale object information. The proposed method is evaluated on two publicly available datasets, namely the WHU-CD and LEVIR-CD datasets. Compared with the state-of-the-art CD methods, the proposed method outperforms in all metrics, with an intersection over union of 92.49% and 85.56%, respectively.

Spatiotemporal fusion (STF) is a cost-effective way to complement the spatiotemporal resolution of multisource images, which has been employed in various applications requiring image sequences. In real-world applications, the spectral accuracy, spatial accuracy, and efficiency of STF play a critical role. Despite this, most STF methods focus on improving spectral accuracy, whereas the challenges of spatial information loss and low efficiency have received limited attention. In addition, the improvements in spectral accuracy, spatial accuracy, and efficiency in STF are contradictory, and existing STF methods cannot balance them well, which limits their reliability and applicability for various STF tasks. To solve the above-mentioned issues, this study proposes an object-level hybrid STF method (OL-HSTFM), which incorporates the efficiency advantage of the object-level fusion strategy, spectral accuracy advantage of the three-step method (Fit-FC), and the spatial accuracy advantage of the spatial and temporal adaptive reflectance fusion model. The performance of OL-HSTFM was compared with two classic STF methods and eight state-of-the-art STF methods at two sites. The experimental results indicate that OL-HSTFM outperforms the other ten methods in overall performance and has excellent efficiency. Furthermore, this study proposes a new metric that can assess the accuracy of both spatial and spectral domains in STF, which provides a more comprehensive and intuitive measurement of the quality of fused images compared to commonly used metrics.

Landslide mapping (LM) from bitemporal remote sensing images is essential for disaster prevention and mitigation. Although bitemporal change detection technology has been applied for LM, there remains room for improvement in its accuracy and automation. In this article, a multilevel feature enhancement network (MFENet) is proposed for LM based on modules built in convolutional neural networks (CNNs) like CNN-Attention. MFENet mainly consists of three modules: the postevent feature enhancement module (PFEM), the bifeature difference enhancement module (BFDEM), and the flow direction calibration module (FDCM). Specifically, the main role of PFEM is to selectively fuse postevent multilayer features to provide discriminative postevent features. BFDEM fuses the multilayer differences of both pre-event and postevent features to generate high-quality change detection features, which are sufficiently powerful to distinguish foreground from background. FDCM uses a digital elevation model to calibrate the flow direction of each pixel of the landslide detection results to complete the LM task. Experiments were conducted to test the effectiveness of MFENet on two real-world regions, Lantau Island and Sharp Peak, Hong Kong, where landslides occurred after rainstorms. Compared with other state-of-the-art general change detection methods and landslide-specific change detection methods, the proposed method outperforms all metrics, with its intersection over union reaching 87.23%. The availability of additional features and the generalization performance of MFENet are demonstrated experimentally. It is anticipated that the proposed network will further contribute to disaster prevention and mitigation.

Object-oriented change detection (OOCD) plays an important role in remote sensing change detection. Generally, most of current OOCD methods adopt the highest predicted probability to determine whether objects have changes. However, it ignores the fact that only parts of an object have changes, which will generate the uncertain classification information. To reduce the classification uncertainty, an improved rough-fuzzy possibilistic $c$-means clustering algorithm combined with multiresolution scales information (MRFPCM) is proposed. First, stacked bitemporal images are segmented using the multiresolution segmentation approach from coarse to fine scale. Second, objects at the coarsest scale are classified into changed, unchanged, and uncertain categories by the proposed MRFPCM. Third, all the changed and unchanged objects in previous scales are combined as training samples to classify the uncertain objects into new changed, unchanged, and uncertain objects. Finally, segmented objects are classified layer by layer based on the MRFPCM until there are no uncertain objects. The MRFPCM method is validated on three datasets with different land change complexity and compared with five widely used change detection methods. The experimental results demonstrate the effectiveness and stability of the proposed approach.

To achieve land cover change detection (LCCD) with both fine spatial and temporal resolutions from remote sensing images, subpixel mapping-based approaches have been widely studied in recent years. The fine spatial but coarse temporal resolution image and the coarse spatial but fine temporal image are used to accomplish LCCD by combining their advantages. However, the performance of subpixel mapping is easily affected by the accuracy of spectral unmixing, thereby reducing the reliability of LCCD. In this article, a novel subpixel change detection scheme based on improved abundance values is proposed to tackle the aforementioned problem, in which the spatial distribution of fine spatial resolution image is borrowed to promote the accuracy of spectral unmixing. First, the coarse spatial resolution image is used to generate the original abundance image by the spectral unmixing method. Second, the spatial distribution information of the fine spatial resolution image is incorporated into the original abundance image to obtain improved abundance values. Third, the fine spatial resolution subpixel map can be generated by the subpixel mapping method using the improved abundance values. At last, the fine resolution change map can be obtained by comparing the subpixel map with the fine spatial resolution image. Experiments are conducted on a simulated dataset based on Landsat-7 images and two real datasets based on Landsat-8 and MODIS images. The results of the real datasets showed that the proposed method can effectively improve the performance of LCCD with an overall accuracy of approximately 1.26% and 0.79% to the existing methods.

Due to its high efficiency and low cost, automatic extraction of building footprints from remotely sensed imagery has long been an important means to obtain building footprint information, which can be easily implemented using existing fully convolutional network (FCN)-based methods. However, such methods suffer from imperfections and thus accurately extracting building footprints from remotely sensed imagery remains a challenging task. For example, cascaded convolutions generally cannot preserve the spatial details well, leading to blurred boundaries and omission of small buildings. Insufficient multiscale features fusion without considering semantic gaps between different level features could yield misclassification. In addition, the fixed receptive field always produces discontinuous holey extracted large buildings. To this end, we propose a novel multiscale and multipath network with boundary enhancement (MMB-Net) that accurately extracts building footprints from remotely sensed imagery. More specially, a parallel multipath feature extraction module is first designed to capture high spatial information-preserved multiscale features with less semantic distances. In addition, the receptive field is enlarged and broadened by a multiscale features enhancement module. Then, an attention-based multiscale features fusion module is built to appropriately aggregate multiscale features. Lastly, a spatial enhancement module is presented to refine the extracted building boundaries by capturing boundary information from low-level features. The proposed MMB-Net has been tested on two benchmark data sets together with other SOTA approaches. The results show that MMB-Net can achieve promising building footprints extraction performances and it outperforms the SOTA methods. The implementation of MMB-Net is available at.

Extracting buildings accurately from very high-resolution (VHR) remote sensing imagery is challenging due to diverse building appearances, spectral variability, and complex background in VHR remote sensing images. Recent studies mainly adopt a variant of the fully convolutional network (FCN) with an encoder–decoder architecture to extract buildings, which has shown promising improvement over conventional methods. However, FCN-based encoder–decoder models still fail to fully utilize the implicit characteristics of building shapes. This adversely affects the accurate localization of building boundaries, which is particularly relevant in building mapping. A contour-guided and local structure-aware encoder–decoder network (CGSANet) is proposed to extract buildings with more accurate boundaries. CGSANet is a multitask network composed of a contour-guided (CG) and a multiregion-guided (MRG) module. The CG module is supervised by a building contour that effectively learns building contour-related spatial features to retain the shape pattern of buildings. The MRG module is deeply supervised by four building regions that further capture multiscale and contextual features of buildings. In addition, a hybrid loss function was designed to improve the structure learning ability of CGSANet. These three improvements benefit each other synergistically to produce high-quality building extraction results. Experimental results on the WHU and NZ32km2 building datasets demonstrate that compared with the tested algorithms, CGSANet can produce more accurate building extraction results and achieve the best intersection over union value 91.55% and 90.02%, respectively. Experiments on the INRIA building dataset further demonstrate the ability for generalization of the proposed framework, indicating great practical potential.

To better interpret the scene and facilitate the subsequent processing of large-scale point cloud, clustering is often implemented in the preprocessing stage. However, when the original density-based spatial clustering of application with noise (DBSCAN) approach is used for point cloud clustering, it is easy to categorize closely spaced vegetation points and nonvegetation points into the same cluster by mistake. Aiming at the problem, this article presents an improved DBSCAN by embedding a strategy of class constraint, which is called CC-DBSCAN. Specially, based on the RGB and label information of each point in the training samples, by using the logistic regression model, the logistic regression color index (LRCI) is calculated for each point in the clustering samples. Then, points to be clustered are classified as vegetation points and nonvegetation points through the LRCI. Furtherly, the class information of point is introduced as a constraint for ensuring the core point and its directly density-reachable points belong to the same class, thus, solving the problem that confusion cluster of the adjacent vegetation points and nonvegetation points. We evaluate our approach on the benchmark SensatUrban dataset, where Cambridge_28 scene dataset is taken as the training set and Cambridge_18 scene dataset is as the dataset to be clustered. Experimental results show that our method achieved 97.20% purity of point cluster, which outperforms the other DBSCAN methods. At the same time, it takes only 24.25 s for clustering 2 million points, which indicates that CC-DBSCAN has high computational efficiency and good practicability.

Urban hotspots are regions with intensive passenger flow, sound infrastructure, and thriving business during a certain period of time, which mirror the travel behavior of residents. Taxi trajectory is one of the important data sources for urban hotspot detection. Unfortunately, it should be pointed out that quite a few of the relevant studies have ignored the temporal dynamics or network-constrained characteristics of urban hotspots, making the detecting results less reasonable and reliable. In this study, a two-phase clustering approach is proposed to detect urban hotspot with taxi trajectory. Concretely, in the first phase, spatiotemporal hierarchical density-based spatial clustering of applications with noise is utilized to cluster the trajectory points with spatial and temporal attributes, which is essential for understanding the evolution of urban hotspots over time. In the second phase, the idea of region growing is introduced to further filter noise, in which the spatial similarity between data points is measured by the route distance, considering that the trajectory data are constrained by the road network. A case study is carried out by the proposed method. Meanwhile, in combination with the Luojia1-01 night-time light remote sensing data and POI data, the reliability of the clustering results is verified and the semantic meaning of the discovered clusters is enriched. Furthermore, not only the spatiotemporal distribution but also the trip lengths and directions of the detected hotspots are explored. These findings can serve as a scientific basis for policymakers in traffic control, public facilities planning, as well as location-based service.

Monitoring pollution of PM2.5 on walkable roads is important for resident health in high-density cities. Due to the spatiotemporal resolution limitations of aerosol optical depth observation, fixed-point monitoring, or traditional mobile measurement instruments, the microscale PM2.5 distribution in the walking environment cannot be fully estimated at the fine scale. In this article, by the integration of mobile measurement data, OpenStreetMap (OSM) data, Landsat images, and other multisource data in land-use regression (LUR) models, a novel framework is proposed to estimate and map PM2.5 distribution in a typical microscale walkable environment of the high-density city Hong Kong. First, the PM2.5 data on the typical walking paths were collected by the handheld mobile measuring instruments, to be selected as the dependent variables. Second, geographic prediction factors calculated by Google Street View, OSM data, Landsat images, and other multisource data were further selected as independent variables. Then, these dependent and independent variables were put into the LUR models to estimate the PM2.5 concentration on sidewalks, footbridges, and footpaths in the microscale walkable environment. The proposed models showed high performance relative to those in similar studies (adj R2, 0.593 to 0.615 [sidewalks]; 0.641 to 0.682 [footpaths]; 0.783 to 0.797 [footbridges]). This article is beneficial for mapping PM2.5 concentration in the microscale walking environment and the identification of hot spots of air pollution, thereby helping people avoid the PM2.5 hotspots and indicating a healthier walking path.

Automatic extraction of road from multisource remote sensing data has always been a challenging task. Factors such as shadow occlusion and multisource data alignment errors prevent current deep learning-based road extraction methods from acquiring road features with high complementarity, redundancy, and crossover. Unlike previous works that capture contexts by multiscale feature fusion, we propose a dual attention dilated-LinkNet (DAD-LinkNet) to adaptively integrate local road features with their global dependencies by joint using satellite image and floating vehicle trajectory data. First, a joint least-squares feature matching-based floating vehicle trajectory correction model is used to correct the floating vehicle trajectory; then a convolutional network model DAD-LinkNet based on a dual-attention mechanism is proposed, and road features are extracted from the channel domain and spatial domain of the target image in turn by constructing a dual-attention module in the dilated convolutional layer and adopting a cascade connection; a weighted hyperparameter loss function is used as the loss function of the model; finally, the road extraction is completed based on the proposed DAD-LinkNet model. Experiments on three datasets show that the proposed DAD-LinkNet model outperforms the state-of-the-art methods in terms of accuracy and connectivity.

Extracting buildings from remotely sensed data is a fundamental task in many geospatial applications. However, this task is resistant to automation due to variability in building shapes and the environmental complexity surrounding buildings. To solve this problem, this article introduces a novel automatic building extraction method that integrates LiDAR data and high spatial resolution imagery using adaptive iterative segmentation and hierarchical overlay analysis based on data fusion. An adaptive iterative segmentation method overcomes over- and undersegmentation based on the globalized probability of boundary contour detection algorithm. A data-fusion-based hierarchical overlay analysis extracts building candidate regions based on segmentation results. A morphological operation optimizes a building candidate region to obtain final building results. Experiments were conducted on the international society for photogrammetry and remote sensing (ISPRS) Vaihingen benchmark dataset. The extracted building footprints were compared with those extracted using the state-of-the-art methods. Evaluation results show that the proposed method achieved the highest area-based quality compared to results from the other tested methods on the ISPRS website. A detailed comparison with four state-of-the-art methods shows that the proposed method requiring no samples achieves competitive extraction results. Furthermore, the proposed method achieved a completeness of 94.1%, a correctness of 90.3%, and a quality of 85.5% over the whole Vaihingen dataset, indicating that the method is robust, with great potential in practical applications.

Car-hailing service has increasingly become popular and fundamentally changed the way people travel in the era of sharing economy. Although such service brings convenience to people's lives, it also causes safety and property concerns. Many studies have been conducted to access the efficiency and effectiveness of car-hailing, but little has been done on its safety monitoring. However, with the rapid development of information technologies such as Internet of Things (IoT), Geographical Information Science (GIS) and automatic monitoring, a more advantageous approach than the current simple drivers screening and testing is feasible. A new model including five indexes i.e. region dangerous index, offset distance of the origin-destination, real-time speed under traffic conditions, vehicle travel time and passenger information, is therefore proposed in this paper based on big data mining of the historical vehicle GPS trajectory data. Experiments were conducted to validate the model in the Gangzha District of Nantong City, China. Several other types of data were used in the experiments, e.g. points of interest (POI), road network data and urban image. The results showed that the proposed model effectively monitored the vehicle when it was driving in a “potentially dangerous area”. In addition, the model could accurately identify the driver's abnormal driving behaviors, such as bypass and abnormal stop. The prediction accuracy of the experiments was 92.06%, among which the discrimination accuracy of the abnormal stop was 100% and that of the detour was 90.57%. All these validate the applicability of the model for future management systems for car-hailing services.

Due to rapid changes on the Earth's surface, it is important to perform land cover change detection (CD) at a fine spatial and fine temporal resolution. However, remote sensing images with both fine spatial and temporal resolutions are commonly not available or, where available, may be expensive to obtain. This paper attempts to achieve fine spatial and temporal resolution land cover CD with a new computer technology based on subpixel mapping (SPM): The fine spatial resolution land cover maps (FRMs) are first predicted through SPM of the coarse spatial but fine temporal resolution images, and then, subpixel resolution CD is performed by comparison of class labels in the SPM results. For the first time, five fast SPM algorithms, including bilinear interpolation, bicubic interpolation, subpixel/pixel spatial attraction model, Kriging, and radial basis function interpolation methods, are proposed for subpixel resolution CD. The auxiliary information from the known FRM on one date is incorporated in SPM of coarse images on other dates to increase the CD accuracy. Based on the five fast SPM algorithms and the availability of the FRM, subpixels for each class are predicted by comparison of the estimated soft class values at the target fine spatial resolution and borrowing information from the FRM. Experiments demonstrate the feasibility of the five SPM algorithms using FRM in subpixel resolution CD. They are fast methods to achieve subpixel resolution CD. [ABSTRACT FROM PUBLISHER]

Object extraction from remote sensing images has long been an intensive research topic in the field of surveying and mapping. Most past methods are devoted to handling just one type of object, and little attention has been paid to improving the computational efficiency. In recent years, level set evolution (LSE) has been shown to be very promising for object extraction in the field of image processing because it can handle topological changes automatically while achieving high accuracy. However, the application of state-of-the-art LSEs is compromised by laborious parameter tuning and expensive computation. In this paper, we proposed two fast LSEs for man-made object extraction from high spatial resolution remote sensing images. We replaced the traditional mean curvature-based regularization term by a Gaussian kernel, and it is mathematically sound to do that. Thus, we can use a larger time step in the numerical scheme to expedite the proposed LSEs. Compared with existing methods, the proposed LSEs are significantly faster. Most importantly, they involve much fewer parameters while achieving better performance. Their advantages over other state-of-the-art approaches have been verified by a range of experiments. [ABSTRACT FROM PUBLISHER]

Indicator cokriging (ICK) has been shown to be an effective subpixel mapping (SPM) algorithm. It is noniterative and involves few parameters. The original ICK-based SPM method, however, requires the semivariogram of land cover classes from prior information, usually in the form of fine spatial resolution training images. In reality, training images are not always available, or laborious work is needed to acquire them. This paper aims to seek spatial structure information for ICK when such prior land cover information is not obtainable. Specifically, the fine spatial resolution semivariogram of each class is estimated by the deconvolution process, taking the coarse spatial resolution semivariogram extracted from the class proportion image as input. The obtained fine spatial resolution semivariogram is then used to estimate class occurrence probability at each subpixel with the ICK method. Experiments demonstrated the feasibility of the proposed ICK with the deconvolution approach. It obtains comparable SPM accuracy to ICK that requires semivariogram estimated from fine spatial resolution training images. The proposed method extends ICK to cases where the prior spatial structure information is unavailable. [ABSTRACT FROM PUBLISHER]

Road information has a fundamental role in modern society. Road extraction from optical satellite images is an economic and efficient way to obtain and update a transportation database. This paper presents an integrated method to extract urban main-road centerlines from satellite optical images. The proposed method has four main steps. First, general adaptive neighborhood is introduced to implement spectral-spatial classification to segment the images into two categories: road and nonroad groups. Second, road groups and homogeneous property, measured by local Geary's C, are fused to improve road-group accuracy. Third, road shape features are used to extract reliable road segments. Finally, local linear kernel smoothing regression is performed to extract smooth road centerlines. Road networks are then generated using tensor voting. The proposed method is tested and subsequently validated using a large set of multispectral high-resolution images. A comparison with several existing methods shows that the proposed method is more suitable for urban main-road centerline extraction. [ABSTRACT FROM PUBLISHER]

There is a type of algorithm for subpixel mapping (SPM), namely, the soft-then-hard SPM (STHSPM) algorithm that first estimates soft attribute values for land cover classes at the subpixel scale level and then allocates classes (i.e., hard attribute values) for subpixels according to the soft attribute values. This paper presents a novel class allocation approach for STHSPM algorithms, which allocates classes in units of class (UOC). First, a visiting order for all classes is predetermined, and the number of subpixels belonging to each class is calculated using coarse fraction data. Then, according to the visiting order, the subpixels belonging to the being visited class are determined by comparing the soft attribute values of this class, and the remaining subpixels are used for the allocation of the next class. The process is terminated when each subpixel is allocated to a class. UOC was tested on three remote sensing images with five STHSPM algorithms: back-propagation neural network, Hopfield neural network, subpixel/pixel spatial attraction model, kriging, and indicator cokriging. UOC was also compared with three existing allocation methods, i.e., linear optimization technique (LOT), sequential assignment in units of subpixel (UOS), and a method that assigns subpixels with highest soft attribute values first (HAVF). Results show that for all STHSPM algorithms, UOC is able to produce higher SPM accuracy than UOS and HAVF; compared with LOT, UOC is able to achieve at least comparable accuracy but needs much less computing time. Hence, UOC provides an effective and real-time class allocation method for STHSPM algorithms. [ABSTRACT FROM AUTHOR]

This paper presents a novel fuzzy-topology-integrated support vector machine (SVM) (FTSVM) classification method for remotely sensed images based on the standard SVM. Induced threshold fuzzy topology is integrated into the standard SVM. First, the optimal intercorrelation coefficient threshold value is applied to decompose an image class in spectral space into the three parts: interior, boundary, and exterior in fuzzy-topology space. The interior-class pixels are then classified as predefined classes based on maximum likelihood. The exterior-class pixels are ignored. The fuzzy-boundary-class pixels which contain misclassified pixels are reclassified based on the fuzzy-topology connectivity theory. As a result, misclassified pixel problems, to a certain extent, are solved. Two different experiments were performed to evaluate the performance of the FTSVM method, in comparison with standard SVM, maximum likelihood classifier (MLC), and fuzzy-topology-integrated MLC. Experimental results indicate that the FTSVM method performs better than the standard SVM and other methods in terms of classification accuracy, hence providing an effective classification method for remotely sensed images. [ABSTRACT FROM AUTHOR]

Object extraction from either aerial or satellite images is a research focus in the field of image processing, where it is essential to determine whether pixels on an image should be classified as object or background. This is a difficult issue, particularly for those pixels which are uncertain for such a determination. This paper presents a newly developed object extraction method that is based on fuzzy topology. First, the idea that the concept object extraction is understood as a fuzzy set in fuzzy topological space is introduced, i.e., the fuzzy object in object extraction, which is composed of its interior and a portion of its boundary, in fuzzy topological space is introduced. Second, the segmentation of the fuzzy object into the interior and the boundary described. Two threshold values induced by the optimal threshold value are proposed for determining the interior, boundary, and exterior of the fuzzy object. Third, object extraction regarding the boundary is introduced, where the connectivity analysis is applied to determine whether the pixels within the boundary should classified as the object. Finally, as an example of application, the developed fuzzy-topology-based object extraction method is applied to boulder extraction from aerial images. The experimental results demonstrate that the newly proposed method provides better object extraction result, particularly to those boundary pixels which are difficult to determine if they should be classified as the object by the existing solutions. The main contribution this paper includes the following two points: 1) Theoretically, this paper introduces fuzzy topology into object extraction where the object and the image background are understood as fuzzy sets in fuzzy topological space, and 2) this paper newly proposes the connectivity analysis method for determining if the pixels within the object boundary should be classified as the object. The overall quality of object extraction is thus improved greatly. [ABSTRACT FROM AUTHOR]

Presents an approach, with emphasis on the proposed line segment matching method, for extracting urban road networks from high-resolution satellite images. Methods for road feature extraction; Characteristics of high-resolution satellite images; Steps to extract road network in the line segment match method.