Accurate traffic forecasting is one of the key applications within Internet of Things (IoT)-based intelligent transportation systems (ITS), playing a vital role in enhancing traffic quality, optimizing public transportation, and planning infrastructure. However, existing spatial-temporal methods encounter two primary limitations: 1) they have difficulty differentiating samples over time and often ignore dependencies among road network nodes at different time scales and 2) they are limited in capturing dynamic spatial correlations with predefined and adaptive graphs. To overcome these limitations, we introduce a novel temporal identity interaction dynamic graph convolutional network (TIIDGCN) for traffic forecasting. The central concept involves assigning temporal identity features to raw data and extracting distinctive, representative spatial-temporal features through multiscale interactive learning. Specifically, we design a multiscale interactive model incorporating both spatial and temporal components. This network aims to explore spatial-temporal patterns at various scales from macro to micro, facilitating their mutual enhancement through positive feedback mechanisms. For the spatial component, we design a new dynamic graph learning method to depict the changing dependencies among nodes. We conduct comprehensive experiments using four real-world traffic datasets (PeMS04/07/08 and NYCTaxi Drop-off/Pick-up). Specifically, TIIDGCN achieves a 16.46% reduction in mean absolute error compared to the Spatial-Temporal Graph Attention Gated Recurrent Transformer Network model on the PeMS08 dataset.

Accurate extraction of aquaculture cages is crucial for effective fishery management. However, this task remains challenging due to high spectral heterogeneity and boundary interference in remote sensing images, which limit the ability of existing methods to automatically generate closed and vectorized cage structures. To address this, we propose VecCageNet, a novel multi-task interactive learning model built on the Segment Anything Model (SAM) for vectorized cage extraction. Our method reformulates cage extraction as a polygon reconstruction problem by identifying valid cage vertices. The model employs a multi-task architecture that simultaneously learns point, edge, and region features using SAM’s powerful representation capabilities. A dedicated feature interaction module integrates these multi-scale features to enhance vertex identification accuracy, enabling the generation of geometrically regular cage polygons. We evaluated VecCageNet on Beijing-3N (BJ-3N) satellite images and compared it with existing semantic segmentation and polygon reconstruction methods. Our method demonstrated superior performance across both pixel-based and object-based accuracy metrics, achieving an IoU of 95.55%. Results justified VecCageNet’s effectiveness in accurately delineating aquaculture net cages with varying geometry and high spectral characteristics, even with limited training samples. Related codes are available at https://github.com/caixiaojiao/VecCageNet.git.

Semantic change detection (SCD) in high-resolution (HR) remote sensing images faces two issues: (1) isolated network branch for binary change detection (BCD) within multi-task architecture result in suboptimal SCD performance; (2) false alarms or missed detections caused by illumination differences or seasonal transform. To address these issues, this study proposes a bi-temporal binary change enhancement network (Bi-BCENet). Specifically, we introduce a binary change enhancement (BCE) strategy based on multi-network joint learning to achieve superior SCD via improving change areas prediction. Within the network’s reasoning process, we develop a cross-attention fusion module (CAFM) to enhance the global similarity modeling via cross-network prompt fusion, and we employ a cosine similarity-based auxiliary loss to optimize non-change’s semantic consistency. The experiments on SECOND and CINA-FX datasets demonstrate that Bi-BCENet outperforms representative SCD networks, achieving 62.08%, 84.95% in FSCD and 66.88%, 83.10% in mIoUsCD, respectively. And the ablation analysis of network validates Bi-BCENet’s effectiveness in reducing false alarms and missed detections in SCD results. Moreover, for specific SCD of cropland, Bi-BCENet shows its strong potential in single-to-multi SCD.

Semantic change detection (SCD) aims to identify potential Earth surface changes, including their location and class, from multi-temporal remote sensing images. However, the under-detection and pseudo-change issues in existing SCD methods severally limit their effectiveness in diverse ground scenarios. To address these issues, a semantic map-guided network, namely SMGNet, is proposed based on a multitask architecture designed to identify potential land cover changes from bi-temporal high-resolution remote sensing images. A robust feature extractor is first developed to extract multi-scale contextual information while retaining fine-grained spatial details, thus enhancing the semantic representation of complex objects with irregular shapes and large sizes. To address the issue of under-detection, we integrate historical semantic information derived from pre-temporal land cover maps into the model using a semantic map encoder module. A semantic fusion module based on Bayesian theory is developed to highlight salient changed information, thus reducing pseudo-changes caused by the same ground objects with spectra variations. Experimental results obtained in a public SCD dataset demonstrate the effectiveness of the proposed method in identifying various semantic changes. Results indicate that the proposed SMGNet achieved the highest detection accuracy, exceeding nine existing methods by 14.81% to 41.28% and 8.45% to 40.31% in terms of SeK andF1scdmetrics on the HRSCD dataset, respectively. The proposed method effectively alleviated pseudo-changes induced by spectra and temporal differences, and accurately detecting these changed objects with irregular shapes and large sizes. The detected results exhibited high inter-class compactness and well-defined boundaries. Code and data are available at https://github.com/long123524/SMGNet.

Change detection in remote sensing images is crucial for assessing human activity impacts and supporting government decision-making. However, in practice, obtaining bi-temporal remote sensing images with consistent conditions is highly limited, and existing change detection methods still face two main challenges: 1) In real-world scenarios, inconsistent sensor and lighting conditions cause significant style (visual appearance) differences between bi-temporal remote sensing images, leading to false changes and reducing change detection accuracy. 2) Remote sensing images contain complex semantic information, and complex scenarios such as shadow occlusion and seasonal vegetation changes make the existing methods difficult to capture relevant features with related to change areas. To address these challenges, we propose a style consistency enhanced differential network (SCEDNet) to eliminate style discrepancies between temporally distinct images and enhance the semantic information of change features. Specifically, we introduce a style consistency module (SCM) in the encoder to extract consistent features by computing the mean and variance of temporal features. Then, we introduce an enhanced differential module (EDM) to enhance change semantics, tackling issues like mislocalization and incomplete regions in complex cases such as shadow occlusion and seasonal vegetation changes. Additionally, we design a gate fusion up-sampling (GFU) and change refine module (CRM) in the decoder to integrate multi-level differential features with different semantic information and highlight key changes, further improving change detection performance. Experiments on the CDD and GZ_CD datasets show that SCEDNet outperformed eight methods, achieving F1-scores of 95.59% and 90.41%, respectively. Code and datasets are available at https://github.com/Yzwfff/SCEDNet.

Semantic change detection (SCD) from very high-resolution images involves two key challenges: (1) the global features of bitemporal images tend to be extracted insufficiently, leading to imprecise land cover semantic classification results, and (2) the detected changed objects exhibit ambiguous boundaries, resulting in low geometric accuracy. To address these two issues, we propose an SCD method called TBSCD-Net based on a multi-task learning framework to simultaneously identify different types of semantic changes and regularize change boundaries. Firstly, we construct a hybrid encoder combining transformer and convolutional neural network (TCEncoder) to enhance the extraction of global context information. A bitemporal semantic linkage module (Bi-SLM) is embedded into the TCEncoder to enhance the semantic correlations between bitemporal images. Secondly, we introduce a boundary-region joint extractor based on Laplacian operators (LOBRE) to regularize the changed objects. We evaluated the effectiveness of the proposed method using the SECOND dataset and a Fuzhou GF-2 SCD dataset (FZ-SCD) and compared it with seven existing methods. The proposed method performed better than the other evaluated methods as it achieved 24.42% Sek and 20.18% GTC on the SECOND dataset and 23.10% Sek and 23.15% GTC on the FZ-SCD dataset. The results of ablation studies on the FZ-SCD dataset also verified the effectiveness of the developed modules for SCD.

Visual foundation models (VFMs) pretrained on large-scale training datasets show robust zero-shot adaptability across many vision tasks. However, there still exist limitations in remote sensing processing tasks due to the variety and complexity of remote sensing images. In this letter, we propose a two-flow network (TFNet) based on multitask VFM, named TFNet, to extract croplands with well-delineated boundaries from high-resolution remote sensing images. TFNet consists of a mask flow and a boundary flow. It first uses a VFM as visual encoder to obtain universal semantic features regarding croplands and then aggregates them into the two flows. Next, a boundary prior-guided module (BPM) is developed to incorporate boundary semantics derived from the boundary flow into the mask flow, to refine the boundary details of croplands. We also develop a multibranch parallel fusion module (MPFM) that aggregates multiscale contextual information to improve the identification of cropland with varied sizes and shapes. Finally, a semantic consistency loss is introduced to further optimize the feature learning of cropland information. We conducted extensive experiments on Shandong (SD) and Xinjiang (XJ) datasets collected from Gaofen-2 (GF-2) satellites and compared our method with five existing methods. Experimental results show that the croplands extracted by our method have the fewest omissions and errors, achieving the highest attribute accuracy (intersection over union (IoU) of 0.863 and 0.945) and lowest geometric errors (global total classification (GTC) of 0.134 and 0.097) than other methods on the two datasets. Our method effectively distinguished croplands of varied sizes, shapes, and spectra, even in scenarios with limited samples. Code and datasets are available athttps://github.com/long123524/TFNet.

Detecting land use changes in urban areas from very-high-resolution (VHR) satellite images presents two primary challenges: 1) traditional methods focus mainly on comparing changes in land cover-related features, which are insufficient for detecting changes in land use and are prone to pseudo-changes caused by illumination differences, seasonal variations, and subtle structural changes and 2) spatial structural information, which is characterized by topological relationships among land cover objects, is crucial for urban land use classification but remains underexplored in change detection. To address these challenges, this study developed a local-global structural interaction network (LGSI-Net) based on a Siamese graph neural network (SGNN) that integrates high-level structural and semantic information to detect urban land use changes from bitemporal VHR images. We developed both local structural feature interaction module (LSIM) and global structural feature interaction module (GSIM) to enhance the representation of bitemporal structural features at the global scene graph and local object node levels. Experiments on the publicly available MtS-WH dataset and two generated datasets, LUCD-FZ and LUCD-HF, show that the proposed method outperforms the existing bag of visual word (BoVW)-based method and CorrFusionNet. Furthermore, we evaluated the detection performance for different semantic feature extraction strategies and structural feature extraction backbones. The results demonstrate that the proposed method, which integrates high-level semantic and graph isomorphism network (GIN)-derived structural features achieves the best performance. The method trained on the LUCD-FZ dataset was successfully transferred to the LUCD-HF dataset with different urban landscapes, indicating its effectiveness in detecting land use changes from VHR satellite images, even in areas with relatively large imbalances between changed and unchanged samples.

This letter presents a cross-learning network (i.e., CLCFormer) integrating fine-grained spatial details within long-range global contexts based upon convolutional neural networks (CNNs) and transformer, for semantic segmentation of very high-resolution (VHR) remote-sensing images. More specifically, CLCFormer comprises two parallel encoders, derived from the CNN and transformer, and a CNN decoder. The encoders are backboned on SwinV2 and EfficientNet-B3, from which the extracted semantic features are aggregated at multiple levels using a bilateral feature fusion module (BiFFM). First, we used attention gate (ATG) modules to enhance feature representation, improving segmentation results for objects with various shapes and sizes. Second, we used an attention residual (ATR) module to refine spatial features’s learning, alleviating boundary blurring of occluded objects. Finally, we developed a new strategy, called auxiliary supervise strategy (ASS), for model optimization to further improve segmentation performance. Our method was tested on the WHU, Inria, and Potsdam datasets, and compared with CNN-based and transformer-based methods. Results showed that our method achieved state-of-the-art performance on the WHU building dataset (92.31% IoU), Inria building dataset (83.71% IoU), and Potsdam dataset (80.27% MIoU). We concluded that CLCFormer is a flexible, robust, and effective method for the semantic segmentation of VHR images. The codes of the proposed model are available athttps://github.com/long123524/CLCFormer.

Two main issues are faced when using very high spatial resolution (VHR) satellite images for building change detection: (1) the boundaries of detected changes are hard to be consistent with the ground truth, and (2) detected changes are easily affected by different viewing angles of bi-temporal images, leading to noticeable false changes. To deal with these issues, this study develops a new Siamese change detection network (i.e., SMCD-Net) based upon a multi-task learning framework to improve building change detection, particularly in the geometric aspect. Boundary information is formulated as an auxiliary task to constrain the learning of high-level semantic features. To enhance the identification of real changes from false changes, we model the directional relationships between buildings and their shadows by fuzzy sets, and incorporate the relationship information into SMCD-Net, leading to a network variant, labeled as SMCD-Net-m. Experiments were conducted on three datasets: a publicly available dataset, a Chinese GaoFen-2 dataset, and a French Pleiades dataset. We compared our methods with seven other methods, i.e., object-based Siamese network, ChangeStar, ChangeFormer, BIT, STANet, FC-Siam-diff, and Siam-NestedUNet. Results showed that the proposed SMCD-Net obtained the best detection results, achieving the lowest global total errors on all datasets. By incorporating directional information, SMCD-Net-m evidently improved detection accuracy, particularly when using bi-temporal images with a large viewing angle difference. The improvement was positively correlated with the accuracy of building shadows extracted from VHR images.

While extracting buildings from high-resolution remote sensing imagery has been widely conducted in automatic surveying and mapping, challenges remain in building extraction over complex scenes, particularly for densely rotated objects with fuzzy boundaries. This study proposes a rotation-aware building instance segmentation network (RotSegNet) that integrates a refined rotated detector to extract rotation equivariant and invariant features. A boundary refinement module is added to the segmentation network to extract fine-grained boundary features. We evaluated our method using the WHU building dataset and AFCities dataset. Our RotSegNet generated a minimum of 2.5% and 0.8% mAP on the two datasets compared with other state-of-the-art methods, which shows the superiority of our method. Results also show that the proposed method can produce regularized buildings with high geometric accuracy.

As part of the G4AW SMARTSeeds project, this study aims to investigate the potential of dense time series of Sentinel-1A dual polarization data for the classification of vegetables that are common in East Java, Indonesia. We first analyzed the temporal behavior of three main types of vegetables (i.e. chili, tomato and cucumber) in terms of backscatter (VH and VV) intensity, and of polarimetric features (i.e. entropy, alpha and anisotropy) derived from polarization decomposition. We then applied a support vector machine with an intersection kernel to the time series data of vegetable samples collected in field. Our results showed that dense time series Sentinel-lA images are of high potential for vegetable classification. Besides using backscatter intensity, the polarimetric information can further improve the discrimination between three specific vegetable types.

This study aims to characterize the spatial arrangement of land cover features and integrate the spatial arrangement information with other commonly used land use indicators. The characterization is conducted at object level, corresponding to land cover objects. At the local urban level, a VHR image is dominated by buildings. Therefore, the characterization of spatial arrangement of land cover elements is mainly conducted on building objects. Since building objects have different functional properties in urban areas, we classify them into a set of different types according to their geometrical, morphological, and contextual properties. The spatial arrangement is characterized by considering the composition of different building types. Regarding the integration of land use indicators and spatial arrangement information, we construct a Bayesian network, in which the spatial arrangement is served as a latent variable, and the land use indicators calculated according to existing studies are treated as the nodes of this Bayesian network. This is followed by urban land use classification. We applied our proposed method to a subset of a Pleiades image for an urban area of Wuhan, China. We conclude that our proposed method can provide an effective means for urban land use extraction.

This paper presents a two-level Active Learning (AL) classification method for the interactive detection of earthquake-induced debris via the synergetic use of post-disaster Very High Resolution (VHR) satellite and local decimeter-resolution aerial images. The proposed method is performed by interactively guiding the human expert in the collection of labeled training samples from aerial images and optimally planning further aerial surveys. A label propagation mechanism is adopted to reliably propagate the labeled pixels annotated by the user on the aerial image by visual photointerpretation to the (lower resolution) satellite image. The experimental analysis is carried out using post-disaster images of the 2010 Yushu earthquake in China. The obtained results confirm that the proposed method can significantly reduce the user annotation effort and the cost for acquiring additional aerial images leading to more accurate and timely debris detection maps.