Search Results (14)

Background/Objectives: It has become a major direction of research in affective computing to explore the brain-information-processing mechanisms based on physiological signals such as electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS). However, existing research has mostly focused on feature- and decision-level fusion, with little investigation into the causal relationship between these two modalities. Methods: In this paper, we propose a novel emotion recognition framework for the simultaneous acquisition of EEG and fNIRS signals. This framework integrates the Granger causality (GC) method and a modality–frequency attention mechanism within a convolutional neural network backbone (MFA-CNN). First, we employed GC to quantify the causal relationships between the EEG and fNIRS signals. This revealed emotional-processing mechanisms from the perspectives of neuro-electrical activity and hemodynamic interactions. Then, we designed a 1D2D-CNN framework that fuses temporal and spatial representations and introduced the MFA module to dynamically allocate weights across modalities and frequency bands. Results: Experimental results demonstrated that the proposed method outperforms strong baselines under both single-modal and multi-modal conditions, showing the effectiveness of causal features in emotion recognition. Conclusions: These findings indicate that combining GC-based cross-modal causal features with modality–frequency attention improves EEG–fNIRS-based emotion recognition and provides a more physiologically interpretable view of emotion-related brain activity. Full article

The Qinba Mountain Region in southern Shaanxi, China, is both a key ecological barrier and a repository of cultural heritage, yet its traditional villages remain highly vulnerable to natural disasters. Disaster-relocation policies have reduced direct exposure to hazards but also created challenges such as settlement hollowing and weakening of cultural continuity. However, systematic studies on the resilience mechanisms of these villages and a corresponding governance framework remain limited. This study applies social–ecological resilience theory to evaluate the resilience of 57 nationally recognized traditional villages. Using a combination of Morphological Spatial Pattern Analysis (MSPA), the entropy weight method, and the geographical detector model, we construct a three-dimensional evaluation framework encompassing terrain adaptability, hazard exposure, and ecological sensitivity. The results show that the Terrain Adaptability Index (TAI) is the dominant driver of resilience, with an explanatory power of q = 0.61, while the interaction of Hazard Exposure Index (HEI, q = 0.58) and Ecological Sensitivity Index (ESI, q = 0.49) produces a nonlinear enhancement effect, significantly increasing vulnerability. Approximately 83% of villages adopt a “peripheral attachment–core avoidance” strategy, and 57% of high-resilience villages (CRI ≥ 0.85) rely on traditional clan-based networks and drainage systems to offset ecological fragility. Based on these differentiated resilience characteristics, the study proposes a three-tiered governance framework of core protection areas–ecological restoration zones–cultural corridors. While this framework demonstrates broad applicability, its findings are context-specific to the Qinba Mountains. Future studies should apply the model to other mountainous regions and integrate dynamic simulation methods to assess climate change impacts, thereby expanding the generalizability of resilience governance strategies. Full article

Emotion analysis based on electroencephalogram (EEG) sensors is pivotal for human–machine interaction yet faces key challenges in spatio-temporal feature fusion and cross-band and brain-region integration from multi-channel sensor-derived signals. This paper proposes MB-MSTFNet, a novel framework for EEG emotion recognition. The model constructs a 3D tensor to encode band–space–time correlations of sensor data, explicitly modeling frequency-domain dynamics and spatial distributions of EEG sensors across brain regions. A multi-scale CNN-Inception module extracts hierarchical spatial features via diverse convolutional kernels and pooling operations, capturing localized sensor activations and global brain network interactions. Bi-directional GRUs (BiGRUs) model temporal dependencies in sensor time-series, adept at capturing long-range dynamic patterns. Multi-head self-attention highlights critical time windows and brain regions by assigning adaptive weights to relevant sensor channels, suppressing noise from non-contributory electrodes. Experiments on the DEAP dataset, containing multi-channel EEG sensor recordings, show that MB-MSTFNet achieves 96.80 ± 0.92% valence accuracy, 98.02 ± 0.76% arousal accuracy for binary classification tasks, and 92.85 ± 1.45% accuracy for four-class classification. Ablation studies validate that feature fusion, bidirectional temporal modeling, and multi-scale mechanisms significantly enhance performance by improving feature complementarity. This sensor-driven framework advances affective computing by integrating spatio-temporal dynamics and multi-band interactions of EEG sensor signals, enabling efficient real-time emotion recognition. Full article

The remote sensing monitoring of land use changes and future scenario simulation hold crucial significance for accurately characterizing urban expansion patterns, optimizing urban land use configurations, and thereby promoting coordinated regional development. Through the integration of multi-source data, this study systematically analyzes the spatiotemporal trajectories and driving forces of land use changes in the Pearl River Delta urban agglomeration (PRD) from 1990 to 2020 and further simulates the spatial patterns of urban land use under diverse development scenarios from 2025 to 2035. The results indicate the following: (1) During 1990–2020, urban expansion in the Pearl River Delta urban agglomeration exhibited a “stepwise growth” pattern, with an annual expansion rate of 3.7%. Regional land use remained dominated by forest (accounting for over 50%), while construction land surged from 6.5% to 21.8% of total land cover. The gravity center trajectory shifted southeastward. Concurrently, cropland fragmentation has intensified, accompanied by deteriorating connectivity of ecological lands. (2) Urban expansion in the PRD arises from synergistic interactions between natural and socioeconomic drivers. The Geographically and Temporally Weighted Regression (GTWR) model revealed that natural constraints—elevation (regression coefficients ranging −0.35 to −0.05) and river network density (−0.47 to −0.15)—exhibited significant spatial heterogeneity. Socioeconomic drivers dominated by year-end paved road area (0.26–0.28) and foreign direct investment (0.03–0.11) emerged as core expansion catalysts. Geographic detector analysis demonstrated pronounced interaction effects: all factor pairs exhibited either two-factor enhancement or nonlinear enhancement effects, with interaction explanatory power surpassing individual factors. (3) Validation of the Patch-generating Land Use Simulation (PLUS) model showed high reliability (Kappa coefficient = 0.9205, overall accuracy = 95.9%). Under the Natural Development Scenario, construction land would exceed the ecological security baseline, causing 408.60 km² of ecological space loss; Under the Ecological Protection Scenario, mandatory control boundaries could reduce cropland and forest loss by 3.04%, albeit with unused land development intensity rising to 24.09%; Under the Economic Development Scenario, cross-city contiguous development zones along the Pearl River Estuary would emerge, with land development intensity peaking in Guangzhou–Foshan and Shenzhen–Dongguan border areas. This study deciphers the spatiotemporal dynamics, driving mechanisms, and scenario outcomes of urban agglomeration expansion, providing critical insights for formulating regionally differentiated policies. Full article

With the rapid deployment of edge base stations and the widespread application of 5G technology, Mobile Edge Computing (MEC)has gradually transitioned from a theoretical concept to practical implementation, playing a key role in emerging human-machine interactions and innovative mobile applications. In the MEC environment, efficiently allocating services, effectively utilizing edge device resources, and ensuring timely service responses have become critical research topics. Existing studies often treat MEC service allocation as an offline strategy, where the real-time location of users is used as input, and static optimization is applied. However, this approach overlooks dynamic factors such as user mobility. To address this limitation, this paper constructs a model based on constraints, optimization objectives, and server connection methods, determines experimental parameters and evaluation metrics, and sets up an experimental framework. We propose an Edge Location Prediction Model (ELPM) suitable for the MEC scenario, which integrates Spatial-Temporal Graph Neural Networks and attention mechanisms. By leveraging attention parameters, ELPM acquires spatio-temporal adaptive weights, enabling accurate location predictions. We also design an improved service allocation strategy, MESDA, based on the Gray Wolf Optimization (GWO) algorithm. MESDA dynamically adjusts its exploration and exploitation components, and introduces a random factor to enhance the algorithm’s ability to determine the direction during later stages. To validate the effectiveness of the proposed methods, we conduct multiple controlled experiments focusing on both location prediction models and service allocation algorithms. The results show that, compared to the baseline methods, our approach achieves improvements of 2.56%, 5.29%, and 2.16% in terms of the average user connection to edge servers, average service deployment cost, and average service allocation execution time, respectively, demonstrating the superiority and feasibility of the proposed methods. Full article

The smart-pixel-based optical convolutional neural network was proposed to improve kernel refresh rates in scalable optical convolutional neural networks (CNNs) by replacing the spatial light modulator with a smart pixel light modulator while preserving benefits such as an unlimited input node size, cascadability, and direct kernel representation. The smart pixel light modulator enhances weight update speed, enabling rapid reconfigurability. Its fast updating capability and memory expand the application scope of scalable optical CNNs, supporting operations like convolution with multiple kernel sets and difference mode. Simplifications using electrical fan-out reduce hardware complexity and costs. An evolution of this system, the smart-pixel-based bidirectional optical CNN, employs a bidirectional architecture and single lens-array optics, achieving a computational throughput of 8.3 × 10¹⁴ MAC/s with a smart pixel light modulator resolution of 3840 × 2160. Further advancements led to the two-mirror-like smart-pixel-based bidirectional optical CNN, which emulates 2n layers using only two physical layers, significantly reducing hardware requirements despite increased time delay. This architecture was demonstrated for solving partial differential equations by leveraging local interactions as a sequence of convolutions. These advancements position smart-pixel-based optical CNNs and their derivatives as promising solutions for future CNN applications. Full article

Unmanned aerial vehicles (UAVs) integrated with computer vision technology have emerged as an effective method for information acquisition in various applications. However, due to the small proportion of target pixels and susceptibility to background interference in multi-angle UAV imaging, missed detections and false results frequently occur. To address this issue, a small target detection algorithm, EDANet, is proposed based on YOLOv8. First, the backbone network is replaced by EfficientNet, which can dynamically explore the network size and the image resolution using a scaling factor. Second, the EC2f feature extraction module is designed to achieve unique coding in different directions through parallel branches. The position information is effectively embedded in the channel attention to enhance the spatial representation ability of features. To mitigate the low utilization of small target pixels, we introduce the DTADH detection module, which facilitates feature fusion via a feature-sharing interactive network. Simultaneously, a task alignment predictor assigns classification and localization tasks. In this way, not only is feature utilization optimized, but also the number of parameters is reduced. Finally, leveraging logic and feature knowledge distillation, we employ binary probability mapping of soft labels and a soft label weighting strategy to enhance the algorithm’s learning capabilities in target classification and localization. Experimental validation on the UAV aerial dataset VisDrone2019 demonstrates that EDANet outperforms existing methods, reducing GFLOPs by 39.3% and improving Map by 4.6%. Full article

Population constitutes the foundational element of cities, and population migration drives the transfer of production factors among urban areas. The population migration network serves as an objective representation of intercity interactions, bearing great significance for the analysis of urban network spatial structure. This study focuses on the 10 core cities within the Poyang Lake urban agglomeration. It utilizes population migration data from Tencent’s location-based big data spanning from 2015 to 2018. Employing the point-axis theory from spatial network theory and the directed weighted network theory within the complex network, the study establishes a comprehensive set of network indices and a network model for spatial structure. It investigates the dynamics of population migration networks within the urban agglomeration and considers strategies for enhancing, regulating, or guiding urban agglomeration development to strengthen its overall vitality. The findings indicate that the urban agglomeration displays distinct characteristics of an urban hierarchical sequence and demonstrates gradual improvement in its spatial network development. While network density remains relatively stable across various threshold intervals over an extended period, network connectivity remains weak. Moreover, the urban agglomeration exhibits the lowest degree of centralization, the highest network structure entropy, and limited network connectivity. Migration along the primary power axis within the urban agglomeration remains relatively stable, while the internal network of the urban agglomeration is interconnected through a “core-non-core” network, reflecting near-geographical connection characteristics. Variations in spatial structure are observed, with the spatial network structure following two modes: “weak core city + edge city” and “node city + outer network city”. The trend in network connections diversifies, resulting in both “core-edge” connections and cross-regional connections. In conclusion, the network characteristics of the urban agglomeration surrounding Poyang Lake are consolidated to aid in formulating an optimization plan for the urban agglomeration’s spatial structure. Additionally, these findings serve as a reference for studying the evolution of spatial structures in the other two urban agglomerations within the city agglomeration in the middle reaches of the Yangtze River. Full article

Expression recognition is a very important direction for computers to understand human emotions and human-computer interaction. However, for 3D data such as video sequences, the complex structure of traditional convolutional neural networks, which stretch the input 3D data into vectors, not only leads to a dimensional explosion, but also fails to retain structural information in 3D space, simultaneously leading to an increase in computational cost and a lower accuracy rate of expression recognition. This paper proposes a video sequence face expression recognition method based on Squeeze-and-Excitation and 3DPCA Network (SE-3DPCANet). The introduction of a 3DPCA algorithm in the convolution layer directly constructs tensor convolution kernels to extract the dynamic expression features of video sequences from the spatial and temporal dimensions, without weighting the convolution kernels of adjacent frames by shared weights. Squeeze-and-Excitation Network is introduced in the feature encoding layer, to automatically learn the weights of local channel features in the tensor features, thus increasing the representation capability of the model and further improving recognition accuracy. The proposed method is validated on three video face expression datasets. Comparisons were made with other common expression recognition methods, achieving higher recognition rates while significantly reducing the time required for training. Full article

In recent years, human action recognition has received increasing attention as a significant function of human–machine interaction. The human skeleton is one of the most effective representations of human actions because it is highly compact and informative. Many recent skeleton-based action recognition methods are based on graph convolutional networks (GCNs) as they preserve the topology of the human skeleton while extracting features. Although many of these methods give impressive results, there are some limitations in robustness, interoperability, and scalability. Furthermore, most of these methods ignore the underlying information of view direction and rely on the model to learn how to adjust the view from training data. In this work, we propose VW-SC3D, a spatial–temporal model with view weighting for skeleton-based action recognition. In brief, our model uses a sparse 3D CNN to extract spatial features for each frame and uses a transformer encoder to obtain temporal information within the frames. Compared to GCN-based methods, our method performs better in extracting spatial–temporal features and is more adaptive to different types of 3D skeleton data. The sparse 3D CNN makes our model more computationally efficient and more flexible. In addition, a learnable view weighting module enhances the robustness of the proposed model against viewpoint changes. A test on two different types of datasets shows a competitive result with SOTA methods, and the performance is even better in view-changing situations. Full article

Vibration dampers and insulators are important components of transmission lines, and it is therefore important for the normal operation of transmission lines to detect defects in these components in a timely manner. In this paper, we provide an automatic detection method for component defects through patrolling inspection by an unmanned aerial vehicle (UAV). We constructed a dataset of vibration dampers and insulators (DVDI) on transmission lines in images obtained by the UAV. It is difficult to detect defects in vibration dampers and insulators from UAV images, as these components and their defective parts are very small parts of the images, and the components vary greatly in terms of their shape and color and are easily confused with the background. In view of this, we use the end-to-end coordinate attention and bidirectional feature pyramid network “you only look once” (BC-YOLO) to detect component defects. To make the network focus on the features of vibration dampers and insulators rather than the complex backgrounds, we added the coordinate attention (CA) module to YOLOv5. CA encodes each channel separately along the vertical and horizontal directions, which allows the attention module to simultaneously capture remote spatial interactions with precise location information and helps the network locate targets of interest more accurately. In the multiscale feature fusion stage, different input features have different resolutions, and their contributions to the fused output features are usually unequal. However, PANet treats each input feature equally and simply sums them up without distinction. In this paper, we replace the original PANet feature fusion framework in YOLOv5 with a bidirectional feature pyramid network (BiFPN). BiFPN introduces learnable weights to learn the importance of different features, which can make the network focus more on the feature mapping that contributes more to the output features. To verify the effectiveness of our method, we conducted a test in DVDI, and its mAP@0.5 reached 89.1%, a value 2.7% higher than for YOLOv5. Full article

Urban science research and the research on megaregions share a common interest in the system of cities and its implications for world urbanization and sustainability. The two lines of inquiry currently remain largely separate efforts. This study aims to bridge urban science and megaregion research by applying network science’s community detection algorithm to explore the spatial pattern of megaregions in the contiguous United States. A network file was constructed consisting of county centroids as nodes, the direct links between each pair of counties as edges, and inter-county commuting flows as the weight to capture spatial interactions. Analyses were carried out at two levels, one at the national level using Gephi and the other for the State of Texas involving NetworkX, an open-source Python programming package to implement a weighted community detection algorithm. Results show the detected communities largely conforming to the qualitative knowledge on megaregions. Despite a number of limitations, the study indicates the great potential of applying network science analytics to improve understanding of the spatial process of megaregions. Full article

Spatial heterogeneity patterns in cities are an essential topic in geographic research and urban planning. This paper analyzes the spatial heterogeneity of places and reflects on the urban structure in cites based on spatial interaction networks. To begin with, we constructed 24 sequentially directed and weighted spatial interaction networks (DWNs) on the basis of points of interest (POIs) and taxi GPS data in Beijing. Then, we merged 24 sequential networks into four clusters: early morning, morning, afternoon, and evening. Next, we introduced the weighted D-core decomposition method in view of the complex network method and weighted distance in a geographic space in order to obtain the in-coreness/out-coreness of places. Finally, three indices (the entropy index, the node symmetry index, and the t-test) were used to measure the heterogeneity of places from both the strength dimension and the direction dimension. The results showed: (1) For the strength dimension, the spatiotemporal strength characteristics of the nodes in the DWN are uneven on weekdays or on the weekends, and the strength heterogeneity on weekdays is more obvious than on weekends; (2) for the direction dimension, out-flows and in-flows are different in the early morning and evening on weekends. In addition, the direction of the DWN is not obvious. The city networks present flat characteristics. This study used the weighted D-core method to identify the heterogeneity of nodes in the DWN, which has certain theoretical and practical value for the planning of urban and urban systems and the coordinated development of cities. Full article

Surface albedo is of crucial interest in land–climate interaction studies, since it is a key parameter that affects the Earth’s radiation budget. The temporal and spatial variation of surface albedo can be retrieved from conventional satellite observations after a series of processes, including atmospheric correction to surface spectral bi-directional reflectance factor (BRF), bi-directional reflectance distribution function (BRDF) modelling using these BRFs, and, where required, narrow-to-broadband albedo conversions. This processing chain introduces errors that can be accumulated and then affect the accuracy of the retrieved albedo products. In this study, the albedo products derived from the multi-angle imaging spectroradiometer (MISR), moderate resolution imaging spectroradiometer (MODIS) and the Copernicus Global Land Service (CGLS), based on the VEGETATION and now the PROBA-V sensors, are compared with albedometer and upscaled in situ measurements from 19 tower sites from the FLUXNET network, surface radiation budget network (SURFRAD) and Baseline Surface Radiation Network (BSRN) networks. The MISR sensor onboard the Terra satellite has 9 cameras at different view angles, which allows a near-simultaneous retrieval of surface albedo. Using a 16-day retrieval algorithm, the MODIS generates the daily albedo products (MCD43A) at a 500-m resolution. The CGLS albedo products are derived from the VEGETATION and PROBA-V, and updated every 10 days using a weighted 30-day window. We describe a newly developed method to derive the two types of albedo, which are directional hemispherical reflectance (DHR) and bi-hemispherical reflectance (BHR), directly from three tower-measured variables of shortwave radiation: downwelling, upwelling and diffuse shortwave radiation. In the validation process, the MISR, MODIS and CGLS-derived albedos (DHR and BHR) are first compared with tower measured albedos, using pixel-to-point analysis, between 2012 to 2016. The tower measured point albedos are then upscaled to coarse-resolution albedos, based on atmospherically corrected BRFs from high-resolution Earth observation (HR-EO) data, alongside MODIS BRDF climatology from a larger area. Then a pixel-to-pixel comparison is performed between DHR and BHR retrieved from coarse-resolution satellite observations and DHR and BHR upscaled from accurate tower measurements. The experimental results are presented on exploring the parameter space associated with land cover type, heterogeneous vs. homogeneous and instantaneous vs. time composite retrievals of surface albedo. Full article