A5101573674- Traffic Prediction and Management Techniques
- Advanced Sensor and Energy Harvesting Materials
- Topic Modeling
- Traffic control and management
- Rough Sets and Fuzzy Logic
- Gas Sensing Nanomaterials and Sensors
- Conducting polymers and applications
- Advanced Image Processing Techniques
- Natural Language Processing Techniques
- Analytical Chemistry and Sensors
- Adaptive Control of Nonlinear Systems
- Advanced Vision and Imaging
- Autonomous Vehicle Technology and Safety
- Sentiment Analysis and Opinion Mining
- Underwater Vehicles and Communication Systems
- Control and Dynamics of Mobile Robots
- Air Quality Monitoring and Forecasting
- Infrastructure Maintenance and Monitoring
- Transportation Planning and Optimization
- Software Engineering Research
- Advanced Fiber Optic Sensors
- Image and Signal Denoising Methods
- Anomaly Detection Techniques and Applications
- Power Systems and Renewable Energy
- Optical measurement and interference techniques
Zhejiang University
2022-2024
Harbin Engineering University
2023
Switch
2023
Beijing University of Posts and Telecommunications
2023
University of Tsukuba
2022-2023
University of Illinois Urbana-Champaign
2022
University of California, Berkeley
2022
Zhejiang University-University of Edinburgh Institute
2021
The EAE task extracts a structured event record from an text. Most existing approaches train the model on each dataset independently and ignore overlap knowledge across datasets. However, insufficient records in single often prevent achieving better performance. In this paper, we clearly define datasets split of into specific target dataset. We propose APE to learn two parts serial learning phases without causing catastrophic forgetting. addition, formulate both as conditional generation...
In dialogue state tracking (DST), the exploitation of history is a crucial research direction, and existing DST models can be divided into two categories: full-history partial-history models. Since “select first, use later” mechanism explicitly filters distracting information being passed to downstream prediction, have recently achieved performance advantage over However, besides redundant information, some critical context was inevitably filtered out by simultaneously. To reconcile...
Queue profile is a crucial measure for traffic management in the vicinity of signalized intersections. In this study, we develop method to identify queue using high resolution data, which can be provided from various sources such as drones. Our methodology has three main components are signal state estimation, identification, and lane detection. The developed algorithms tested on real-world dataset collected by drones case study validation. Remarkably, our only uses drone data input it...
Short-term traffic prediction is one of the most important elements in Intelligent Transportation Systems (ITS). Although an extensive collection state-of-the-art methods emerge, well-established approaches require stringent assumptions that both training data and testing must be same feature space with independent identical distribution sufficient representative data, including features space. However, due to systematic measurement error, communication problems, device failure, sensors...
The quality of three-dimensional (3D) reconstruction algorithm Structure from Motion (SfM) is affected by the input image's resolution and CMOS's noise level. We propose a denoisable Super Resolution (SR) method to improve resolutions while denoising for SfM's images taken CMOS device, improving its performance on noisy images. conventional deep learning SR does not consider during process. This results in disability simultaneously reducing resolution. In our methods Add Noise before...
The resolutions and noise levels of input images directly affect three-dimensional (3D) structure-from-motion (SfM) reconstruction performance. Conventional super-resolution (SR) methods focus too little on denoising; moreover, latent image often worsens when improving resolution. This paper proposes two SR denoising training algorithms to improve both resolution simultaneously: add-noise-before-downsampling downsample-before-adding-noise. These preprocess low-resolution using real-world...
Short-term traffic forecasting has been a hot topic in the intelligent transportation systems field. The traditional methods mostly fix sensors. However, most sensors are subject to bad conditions, leading noisy and insufficient raw data. Recent advances have provided new prediction opportunities. For example, transfer learning method takes advantage of data trained on one good dataset transfers knowledge others with Existing applications do not consider underlying distributions...