A5040998859- Anomaly Detection Techniques and Applications
- Fuel Cells and Related Materials
- Electrocatalysts for Energy Conversion
- Video Surveillance and Tracking Methods
- Human Pose and Action Recognition
- Advanced Image and Video Retrieval Techniques
- 3D Surveying and Cultural Heritage
- Multimodal Machine Learning Applications
- Sparse and Compressive Sensing Techniques
- Robotic Path Planning Algorithms
- Domain Adaptation and Few-Shot Learning
- Autonomous Vehicle Technology and Safety
- Membrane-based Ion Separation Techniques
- Time Series Analysis and Forecasting
- Advanced Multi-Objective Optimization Algorithms
- Reinforcement Learning in Robotics
- Spectroscopy and Chemometric Analyses
- Remote-Sensing Image Classification
- Advanced Vision and Imaging
- Image Processing Techniques and Applications
- Advanced battery technologies research
- Distributed Control Multi-Agent Systems
- Video Analysis and Summarization
- Advanced Measurement and Detection Methods
- Advanced Statistical Methods and Models
University of Science and Technology of China
2021-2025
China General Nuclear Power Corporation (China)
2024
Nanyang Technological University
2023
Shenzhen University
2023
Dalian National Laboratory for Clean Energy
2022
Hefei National Center for Physical Sciences at Nanoscale
2022
Anhui University
2022
SRI International
2017-2018
China University of Geosciences
2008
Abstract Under the growing crisis of coronavirus disease 2019 pandemic, global medical system is facing predicament an acute shortage medical‐grade oxygen (O 2 , ≥ 99.5% purity). Herein, generation device manufactured that relies on electrochemical technology. The performance generator (EOG) remarkably improved to a practically applicable level, achieving long‐term (>200 h), stable, and quick production (>1.5 L min −1 ) high purity O (99.9%) at energy efficiency (496 kW h ), via...
Video data are distinct from images for the extra temporal dimension, which results in more content dependencies various perspectives. It increases difficulty of learning representation video actions. Existing methods mainly focus on dependency under a specific perspective, cannot facilitate categorization complex This paper proposes novel selective aggregation (SDA) module, adaptively exploits multiple types to refine features. Specifically, we empirically investigate long-range and...
Although electrochemical technologies offer vast industrial prospects, broader adoption-particularly in consumer applications-remains constrained by high costs and limited component lifespans. Here, we present a gravity-assisted, membrane-free oxygen (O2) removal (EOR) reactor coupling reduction (ORR) evolution (OER) reactions. Leveraging fluid mechanics insights, buoyant O2 bubbles ascend rapidly, achieving 95% product self-separation eliminating the need for membranes or external...
Efficient action recognition aims to classify a video clip into specific category with low computational cost. It is challenging since the integrated spatial-temporal calculation (e. g., 3D convolution) introduces intensive operations and increases complexity. This paper explores feasibility of integration channel splitting filter decoupling for efficient architecture design feature refinement by proposing novel spatio-temporal collaborative (STC) module. STC splits channels two groups...
In this paper, we propose a multi-agent reinforcement learning approach, POCA-Mix, to achieve collaborative multi-target search with visual drone swarm. The proposed approach leverages the benefits of curriculum and mixed credit assignment guide swarm in per-forming task only local perception constrained 3D environment. To validate performance conducted simulation experiments various combinations regarding number drones targets. results demonstrate that outperforms other baseline methods...
This paper describes a radar-guided monocular vision system that detects, validates, and tracks the preceding vehicle thus predicts its lane-change intentions. A vision-based lane tracking process is developed to create stable motion model in order map radar targets image coordinates consequently generate region of interest (ROI) search for potential vehicle. Model-based object classification algorithms are then applied validate existence this ROI. Once detected primary target vehicle, which...
We present a novel method for vehicle detection and classification in aerial imagery. First, change analyzes pair of mutually aligned images captured at the same location but different time instants to generate proposals. Next, trained Convolutional Neural Network (CNN) classifier is applied (1) determine if proposal truly contains vehicle, (2) classify any into major categories. Experimental results using infrared (IR) data demonstrate efficacy our method: rate over 99%; light-duty vehicles...
Videos naturally contain dynamic variation over the temporal axis, which will result in same visual clues (e.g., semantics, objects) changing their scale, position, and perspective patterns between adjacent frames. A primary trend video CNN is adopting spatial-2D convolution for spatial semantics temporal-1D dynamics. Though direction achieves a favorable balance efficiency efficacy, it suffers from misalignment of with large displacements. Particularly, rigid would fail to capture correct...
This paper presents a real time system for tracking multiple ground moving targets in aerial video. The state of target is described by its kinematics as well shape and appearance features: the include location velocity an earth fixed coordinate system; parameters ellipse; features consist color histogram, correlogram, edge matching and/or orientation correlation information. represented constant model are static models between two observation instances. motion layers elliptical shapes...
Vision-language models, such as CLIP, have shown impressive generalization capacities when using appropriate text descriptions. While optimizing prompts on downstream labeled data has proven effective in improving performance, these methods entail labor costs for annotations and are limited by their quality. Additionally, since CLIP is pre-trained highly imbalanced Web-scale data, it suffers from inherent label bias that leads to suboptimal performance. To tackle the above challenges, we...
Abstract Under the growing crisis of coronavirus disease 2019 pandemic, global medical system is facing predicament an acute shortage medical-grade oxygen (O 2 , ≥ 99.5% purity). Herein, we manufactured generation device relying on electrochemical technology. The performance generator (EOG) was remarkably improved to a practically applicable level, achieving long-term (>200 h), stable, and quick production (>1.5L/min) high purity O (99.9%) under energy efficiency (496 L/kW·h),...
Change detection in aerial imagery is challenging due to the effects of platform movements, atmospheric degradation and unpredictable photometric or thermal variations scene. We introduce an change system built around multiple detectors, each exploiting a different feature set, fusion their output order overcome challenges. A supervised classifier screens fused results for false detections. In particular, we have investigated three detectors: Histogram Oriented Gradients (HOG),...
Abstract Image restoration is a widely studied problem in the field of image processing. Although existing methods based on denoising regularization have shown relatively well performance, for different features unknown images not been proposed. Since features, it seems necessary to adopt priori regular terms features. In this paper, we propose multiscale framework that can simultaneously perform two or more prior better obtain overall results. We use alternating direction multiplier method...