Tao Wu

Deploying deep learning models for high-resolution remote sensing image segmentation remains challenging in resource-constrained scenarios due to the high computational cost of dense prediction and the structural vulnerability of thin objects such as roads. To address these challenges, we propose LiteRoadSegNet, a lightweight and deployment-oriented segmentation framework that achieves a favorable balance among efficiency, accuracy, and structural preservation. The proposed model adopts a compact encoder–decoder architecture composed of a lightweight hierarchical vision transformer and a streamlined decoder, enabling efficient multi-scale feature representation under limited computational budgets. To enhance structural consistency without increasing inference overhead, we further design a low-cost semantic–topological dual-branch contrastive learning scheme which enhances feature discriminability and preserves road connectivity during training. In addition, to improve deployment robustness in cross-region scenarios, we incorporate a lightweight test-time adaptation strategy based on Adaptive Batch Normalization (AdaBN) and sliding-window inference. This strategy enables seamless adaptation to unlabeled target domains without requiring model retraining. Extensive experiments demonstrate that LiteRoadSegNet achieves competitive segmentation performance and superior topology preservation while maintaining a small model footprint and high inference efficiency, making it well suited for large-scale remote sensing applications under resource-constrained environments.

Personalized urban services are becoming increasingly significant in smart city systems. This shift from intelligent transportation to smart cities broadens the scope of personalized services, encompassing not just travel but a wide range of urban activities and needs. This study proposes a knowledge graph-based Hidden Markov Model (KHMM) to improve personalized route recommendations by incorporating both spatial and semantic relationships between Points of Interest (POIs) in a unified decision-making framework. The KHMM expands the state space of the traditional Hidden Markov Model using a knowledge graph, enabling the integration of multi-dimensional POI information and higher-order relationships. This approach reflects the spatial complexity of urban environments while addressing user-specific preferences. The model’s empirical evaluation, focused on Changsha, China, examined how temporal variations in public attention to POIs influence route selection. The results show that incorporating dynamic temporal and spatial data significantly enhances the model’s adaptability to changing user behaviors, supporting real-time, personalized route recommendations. By bridging individual preferences and road network structures, this research provides key insights into the factors shaping travel behavior and contributes to the development of adaptive and responsive urban transportation systems. These findings highlight the potential of the KHMM to advance intelligent travel services, offering improved spatial accuracy and personalized route planning.

The Segment Anything Model (SAM) enables zeroshot or interactive segmentation of visual content and has shown remarkable performance with natural images. However, SAM often fails to accurately extract roads from high-resolution remote sensing imagery (HRSI) due to complex road boundaries, background interference, and its semantic-agnostic nature. This paper proposes an automated road extraction method called Trajectory Points Prompting SAM (TPP-SAM), which uses trajectory points as segmentation prompts for roads in HRSI, eliminating the need for complex manual prompts. TPPSAM introduces road centerline constraint (RCC) and sampling constraint (SC) layers to select reliable trajectory prompt points and migrates centroid information from building roofs as a background constraint (BC). This three-layer constraint system integrates real-world crowdsourced information with the vision foundation model. Additionally, the data structure of prompt points is organized using map matching and recoding rules, accounting for their attributes and relative positional coordinates. The processed data are then embedded into the prompt encoder to generate road semantic masks. Results show that TPPSAM achieves comprehensive road segmentation across various scenarios, demonstrating the viability of using trajectory points as interactive road segmentation prompts. The proposed model extends the application of geographic image segmentation and holds potential for other geolabeled features.

The relationship between commercial vehicle travel patterns and urban functional areas reveals potential connections between urban form and human geographic flows, which provides critical information for optimizing urban transportation systems. Benefiting from the large-scale trajectory datasets, it would be possible to investigate deeper research by modeling the implied urban travel patterns. This study designs a framework to reveal the collective movement patterns of commercial vehicle trajectories inside the urban environment, focusing on their spatiotemporal variations within functional areas. Stopping behaviors of trajectories were identified to construct spatiotemporal origin-destination (OD) matrices, representing time-varying human geographic flows. The singular value decomposition (SVD) method was employed to quantify spatio-temporal OD matrice to obtain time and space travel features. Travel patterns’ dynamics and spatial interactions within functional areas were then analyzed. The experimental results obtained with real-life datasets from Changsha, China, uncovered three typical travel patterns depicting commercial vehicle activities in urban environment shifts from work-related locations on weekdays to leisure destinations on weekends, with central areas experiencing more short and medium-range trips. The findings provide scientific references for optimizing spatio-temporal travel patterns and functional distribution to meet the demands of urban development and traffic management strategies.

With the rapid development of transport technology and the increasing complexity of traffic patterns, integrating multiple data sources for traffic flow prediction has become crucial to overcome the defects of a single data source. This paper introduces a multisource data fusion approach with graph convolutional neural networks (GCNs) for node‐level traffic flow prediction. Specifically, it extracts different types of traffic flows from multiple data sources and constructs a unified graph structure by using global traffic nodes to interpolate the traffic flow. In addition, a GCN combined with gated recurrent units (GRUs) is proposed for spatiotemporal modeling of data fusion and traffic flow prediction. The main contributions are: (1) The approach significantly improved prediction accuracy by leveraging multiple data sources compared to a single source. (2) A unified graph structure was created via global traffic nodes to interpolate traffic flow and address data sparsity. (3) The proposed model demonstrates an over 11% improvement in accuracy compared to other baseline models, as measured by the weighted mean absolute percentage error (WMAPE). It also exhibits stability in multitime scale predictions, highlighting the effectiveness of multisource data fusion, data imputation, and node‐level prediction capabilities. The approach provides valuable insights for managing urban traffic data from multiple sources and predicting traffic flow, and it shows stability in multitime scale predictions.

Lane-level road information is especially crucial now that high-precision navigation maps are in more demand. Road information may be obtained rapidly and affordably by mining floating vehicle data (FCD). A method is proposed to extract the number of lanes on urban roads by combining deep learning and low-frequency FCD. Initially, the FCD is cleaned using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering technique. Then, the FCD is split into three categories based on the typical urban road types: one-way one-lane, one-way two-lane, and one-way three-lane, and the deep learning sample data is created using segmentation, rotation, and gridding. Lastly, the number of urban road lanes is obtained by training and predicting the sample data using the LeNet-5 model. The number of urban road lanes was effectively identified from the low-frequency FCD with a detection accuracy of 92.7% through the cleaning and classification of Wuhan FCD. Urban roads can be efficiently covered by the FCD on a regular basis, and lane information can be efficiently collected using deep learning techniques. This method can be used to generate and update lane number information for high-precision navigation maps.