A5100448645- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Human Pose and Action Recognition
- Caching and Content Delivery
- Anomaly Detection Techniques and Applications
- Advanced Data Storage Technologies
- Machine Learning and Data Classification
- Video Surveillance and Tracking Methods
- Face and Expression Recognition
- Advanced Database Systems and Queries
- Image Retrieval and Classification Techniques
- Cloud Computing and Resource Management
- Advanced Clustering Algorithms Research
- Data Management and Algorithms
- Robotics and Sensor-Based Localization
- Vehicle License Plate Recognition
- Web Data Mining and Analysis
- Robotic Path Planning Algorithms
- Adversarial Robustness in Machine Learning
- Spam and Phishing Detection
- Power Line Inspection Robots
- Complex Network Analysis Techniques
- Power Line Communications and Noise
- VLSI and FPGA Design Techniques
- Magnetic properties of thin films
Beijing Microelectronics Technology Institute
2020-2025
Wuhan University of Technology
2024-2025
Chinese Academy of Social Sciences
2023-2025
Institute of Archaeology
2023-2025
Dalian University
2021-2024
Beihang University
2013-2024
University of Chicago
2020-2024
University of Alberta
2024
Energy Research Institute
2024
Xi’an University of Posts and Telecommunications
2017-2024
Existing substation equipment image data augmentation models face challenges such as high dataset size requirements, difficult training processes, and insufficient condition control. This paper proposes a transformer method based on Stable Diffusion model. The proposed incorporates the Low-Rank Adaptation (LoRA) concept to fine-tune pre-trained model weights, significantly reducing requirements while effectively integrating essential features of data. To minimize interference from complex...
The unearthed artifacts from the ruins of Weiyang Palace in Han Dynasty's Chang'an City include a large number bone stick relics. front and back each form pair, with similar appearances within indicating matching relationship. However, close resemblance shape complex texture features sticks make traditional image methods inefficient less accurate. To address these challenges, this paper proposes ConvNeXt-based Siamese network for matching, utilizing feature similarity metrics.The method...
TEB (timed elastic band) can efficiently generate optimal trajectories that match the motion characteristics of car-like robots. However, quality generated is often unstable, and they sometimes violate boundary conditions. Therefore, this paper proposes a fuzzy logic control–TEB algorithm (FLC-TEB). This method adds smoothness jerk objectives to make trajectory by smoother control more stable. Building on this, controller proposed based kinematic constraints It uses narrowness turning...
The control plane is decoupled from data in the Software- Defined Networking (SDN) architecture, under which network intelligence and state are logically centralized underlying infrastructure abstracted applications. Load balancing a significant technology it can help save power improve resource utilization network. Since SDN makes possible to manage whole with controller, we distribute traffic more efficiently easily. In this paper, research background of OpenFlow-based summarized firstly....
Clustering algorithm is widely used in data mining. It attempt to classify elements into several clusters, and the same cluster are more similar each other meanwhile belonging clusters not similar. The recently published density peaks clustering can overcome disadvantage of distance-based that only find nearly-circular shapes, instead it discover arbitrary shapes insensitive noise data. However needs calculate distances between all pairs points scalable big data, order reduce computational...
Transmission tower re-identification refers to the recognition of location and identity transmission towers, facilitating rapid localization towers during power system inspection. Although there are established methods for defect detection accessories (such as crossarms insulators), is a lack automated matching. This paper proposes an identity-matching method that integrates machine vision deep learning. Initially, requires creation template library. Firstly, YOLOv8 object algorithm employed...
Accurate citation count prediction of newly published papers could help editors and readers rapidly figure out the influential in future. Though many approaches are proposed to predict a paper's future citation, most ignore dynamic heterogeneous graph structure or node importance academic networks. To cope with this problem, we propose Dynamic Graph Node Importance network (DGNI) learning framework, which fully leverages information trends papers. First, embedding module is provided capture...
Intelligent robotic assistance requires a robot to accurately understand human behavior. Many researchers have explored human-object interactions decode behavior-related information. However, current methods only model probabilistic correlations between objects and activities. Their applications are usually limited fixed environments sets of They unable deal with variability in the real due lack human-like cognitive reasoning process. To address this urgent problem, we developed an Object...
The detection of the safety helmet is difficulties due to ariable ighting, weather changes and complex background. We proposed a deep learning method detect solve problems low accuracy poor robustness traditional methods. This based on SSD (Single Shot MultiBox Detector) object improved network. First, we used fusion multi-layer consideration shallow sematic information semantic information, which improves sensitivity network small target detection. Second, lightweight structure compresses...
The Fuzzy C-means (FCM) algorithm is one of the most widely used algorithms in unsupervised pattern recognition. As intensity observation noise increases, FCM tends to produce large center deviations and even overlap clustering problems. relative entropy fuzzy (REFCM) adds as a regularization function algorithm, which has good ability for detection membership assignment observed values. However, REFCM still generate overlapping clusters size cluster increases becomes imbalanced. Moreover,...