A5011290020- Video Surveillance and Tracking Methods
- Traffic Prediction and Management Techniques
- Time Series Analysis and Forecasting
- Advanced Neural Network Applications
- Traffic control and management
- Image Enhancement Techniques
- Advanced Vision and Imaging
- Human Mobility and Location-Based Analysis
- Korean Urban and Social Studies
- Simulation and Modeling Applications
- Data Visualization and Analytics
- Robotic Path Planning Algorithms
- Advanced Computational Techniques and Applications
- Face and Expression Recognition
- Advanced Image and Video Retrieval Techniques
- Advanced Image Processing Techniques
- Industrial Vision Systems and Defect Detection
- Autonomous Vehicle Technology and Safety
- Automated Road and Building Extraction
- Image and Object Detection Techniques
- Transportation and Mobility Innovations
- Color Science and Applications
- Anomaly Detection Techniques and Applications
- Diverse Aspects of Tourism Research
Hangzhou Dianzi University
2017-2025
The existence of specular highlights hinders high-level computer algorithms. In this paper, we propose a novel approach to remove from single grayscale image by regarding the problem as an image-to-image translation task between highlight domain and diffuse domain. We solve using generative adversarial network framework, where generator removes discriminator judges whether outputs are clear highlight-free. Specular removal is intractable should while keeping many details possible....
The transportation detection of long-distance small objects has low accuracy. In this work, we propose DMF, which is based on disparity depths. We map different regions to 2D candidate according the distance solve small-object problem. This method clusters maps projected image extracted with features in mapping region. On one hand, it uses a multicluster unsample regions. other feature fusion scales performed each cluster experimental results two datasets show that DMF can improve accuracy objects.
Traffic flow prediction methods commonly rely on historical traffic data, such as volume and speed, but may not be suitable for high-capacity expressways or during peak hours. Furthermore, downstream can have significant impacts flow. To address these challenges, our study proposes a novel model, V-STF, which integrates visual to quantify macroscopic indicators, well density features in temporal feedback spatio features. The contribution of proposed model lies its ability improve accuracy...
In an industrial environment, measuring and reconstructing metal objects using computer vision methods can be affected by surface highlight reflections, leading to inaccurate results. this article, we propose a novel network with broad applicability for removing reflections based on dynamic masks, which is suitable images where baseline image no not available. First, use pretrained model learn the features of surface, then construct adaptive soft mask that allows texture in regions preserved...
Accurate traffic flow prediction is essential for applications like transport logistics but remains challenging due to complex spatio-temporal correlations and non-linear patterns. Existing methods often model spatial temporal dependencies separately, failing effectively fuse them. To overcome this limitation, the Dynamic Spatial-Temporal Trend Transformer DST2former proposed capture through adaptive embedding dynamic static information learning multi-view features of networks. The approach...
An understanding of the evolutionary patterns in areas urban activity is crucial for official decision makers and planners. The origin-destination (OD) datasets generated by human daily travel behavior reflect dynamics. Previous spatio-temporal analysis methods utilize these to extract popular city areas, with ignorance flow relationships between areas. Several have been unable determine time steps similar spatial characteristics automatically or failed recognize various modalities a city....
The stereo vision could obtain the 3-D coordinate of detected object by computing disparity corresponding image points. However, on account time complexity and low robustness matching algorithm, it is seldom used in large-scale scene. This paper puts forward a new vehicle detection method, which simplifies massive Fourier transformation process. method converts 2-D to 1-D with dimensionality reduction reused transformation. Meanwhile, fast model also derived. coarse-to-fine pyramid search...
Detection methods based on 2-D images tend to extract the color, texture, shape, and other appearance features of objects. However, in complex scenes, detection results using these are often influenced by shadows, occlusion, resolution. In this paper, a disparity-proposal-based method that rapidly extracts candidate frames objects basis stereo disparity ensures robustness under different perturbations is proposed. Furthermore, depth information used construct multi-scale pooling layers,...