A5107005798- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Image Retrieval and Classification Techniques
- Advanced Image and Video Retrieval Techniques
- Image and Signal Denoising Methods
- Advanced Vision and Imaging
- Advanced Algorithms and Applications
- Remote Sensing and Land Use
- Banking stability, regulation, efficiency
- Image Processing Techniques and Applications
- Infrared Target Detection Methodologies
- Advanced Measurement and Detection Methods
- Simulation and Modeling Applications
- Fire Detection and Safety Systems
- Firm Innovation and Growth
- Advanced Image Processing Techniques
- Medical Image Segmentation Techniques
- UAV Applications and Optimization
- Microgrid Control and Optimization
- Advanced Image Fusion Techniques
- Image Enhancement Techniques
- Visual Attention and Saliency Detection
- Privacy-Preserving Technologies in Data
- Sports Dynamics and Biomechanics
- CCD and CMOS Imaging Sensors
China University of Mining and Technology
2024
Kunming University of Science and Technology
2020-2024
Kunming University
2024
Tsinghua University
1998-2024
Tongji University
2023-2024
Yanshan University
2024
Beijing Electronic Science and Technology Institute
2023-2024
Zhongda Hospital Southeast University
2023
Wuhu Hit Robot Technology Research Institute
2022
Beijing University of Civil Engineering and Architecture
2021-2022
In the field of remote sensing image applications, RGB and infrared object detection is an important technology. The performance can be improved robustness algorithm will enhanced by making full use their complementary information. Existing RGB-infrared methods do not explicitly encourage images to achieve effective multimodal learning. We find that when fusing images, cross-modal redundant information weakens degree fusion. Inspired this observation, we propose a suppression network...
Aerial object tracking has recently shown great potential in the field of remote sensing. However, small objects with limited feature information pose a huge challenge to aerial trackers. Despite significant improvements, most trackers still struggle capture enough discriminative features and overcome background disturbances. In this work, we propose an efficient tracker (SmallTrack) based on Siamese network improve discrimination objects. It consists two effective modules, namely Wavelet...
Abstract This study aims to promote the protection and inheritance of cultural heritage in private gardens Jiangnan area China. By establishing a precise visual labeling system accelerating construction database for garden features, we deepen understanding design philosophy. To this end, propose an improved recognition model based on You Only Look Once (YOLO) v8. is particularly suitable processing environments with characteristics such as single or complex structures, rich depth field,...
To leverage enormous unlabeled data on distributed edge devices, we formulate a new problem in federated learning called Federated Unsupervised Representation Learning (FURL) to learn common representation model without supervision while preserving privacy. FURL poses two challenges: (1) distribution shift (Non-IID distribution) among clients would make local models focus different categories, leading the inconsistency of spaces. (2) unified information FURL, representations across be...
Abstract The convergence of cultural and aesthetic elements in timber structures from China’s Tang Dynasty (618–907 AD) traditional Japanese architecture provides a rich tapestry architectural evolution cross-cultural exchanges. Addressing the challenge distinguishing understanding intricate styles these is significant for both historical comprehension preservation efforts. This research introduces an innovative approach by integrating Multi-Head Attention (MHA) mechanism into YOLOv8 model,...
Text detection and recognition in natural scene images plays an important role content analysis of images. In this paper, based on the characteristics text, we propose a robust text method using Maximally Stable Extremal Regions (MSER) Support Vector Machine (SVM). Different from end to recognition, split problem into procedure. Firstly, stage, order extract potential as much possible, use MSER color clustering connected component. Then, for obtained candidate component, visual saliency some...
A dynamic model of gear transmission system wind turbine is built with consideration randomness loads and parameters. The response the obtained using theory random sampling Runge-Kutta method. According to rain flow counting principle, meshing forces are converted into a series luffing fatigue load spectra. amplitude frequency equivalent stress method Geber quadratic curve. Moreover, reliability components according probability cumulative damage. variation parameters calculated influence on...
Abstract Small object detection has been a problem in deep learning convolutional neural network models. A multi-rate dilated convolution module is proposed to form feature map locate small objects. Benefiting from the fact that does not add extra complexity while maintaining characteristics of high-resolution map, this paper replaces traditional by setting with ratio 1, 2, and 5, fuses different rate convolutions. The branch, extracting high-level features, using FPN algorithm as basic...
Over an extended period, considerable research has focused on elaborated mapping in navigation systems. Multispectral point clouds containing both spatial and spectral information play a crucial role remote sensing by enabling more accurate land cover classification the creation of maps. However, existing graph-based methods often overlook individual characteristics patterns these graphs, leading to convoluted pattern aggregation failure fully exploit spatial–spectral classify multispectral...
This paper presents an improvement in power flow calculation based on current injection method by introducing optimization factor. In the proposed this paper, PQ buses are represented mismatches while PV mismatches. It is different from representations conventional equations. By using combined and method, number of equations required can be decreased to only one for each bus. The factor used improve iteration process ensure effectiveness improved when system ill-conditioned. To verify IEEE...
Characterized by operating regionally, "credit only, non deposit", loans in small, scattered and short-term, micro-loan companies have became a major force of financial services rural areas China increasingly. Due to differences the lending object basis loan decision, credit risk evaluation large amounts business are completely different, current methods can't be applicable absolutely. Bases on review foreign countries in-depth analysis special reasons that induce companies, we found...
Existing RGB-Infrared detection models do not explicitly encourage RGB and infrared to achieve effective multimodal learning. We find that when fusing images, cross-modal redundant information weakens the degree of complementary fusion. Inspired by this observation, we propose Redundant Information Suppression Network (RISNet) which suppresses facilitates fusion information. Specifically, design a novel mutual minimization module reduce redundancy between appearance features from images...
High-precision industrial manipulators are essential components in advanced manufacturing. Model-based feedforward is the key to realizing high-precision control of robot manipulators. However, traditional approaches based on rigid models or flexible joint which neglect elasticities out rotational directions and degrade setpoint precision significantly. To eliminate effects all directions, a method proposed output redefinition extended model (EFJM). First, robots modeled by EFJM describe...
To study the electromagnetic (EM) backscatter characteristics of freak waves at moderate incidence angles, we establish an EM backscattering model for in (1 + 1)-dimensional deep water. The nonlinear interaction between and Bragg short is considered to be basic hydrodynamic spectra modulation mechanism model. Numerical results suggest that intensities are less than those from background sea surface angles. normalised radar cross sections (NRCSs) highly polarisation dependent, even low which...
We propose a novel single image super-resolution (SR) algorithm based on the projective dictionary pair learning with anchored neighborhood regression. Different from previous methods that aim to learn only synthesis or an analysis dictionary, our method would both types of dictionaries jointly for regression achieve SR. first cluster training features into K clusters in order and dictionaries. Moreover, we regressions samples at phase use them reconstruction stage. As shown experimental...
We examine the impact of intellectual property rights (IPR) protection on local private firm innovation. Using establishment patent courts in China as exogenous shocks, we show that patenting activities cities with are significantly higher than those without courts. Consistent our hypotheses, effect is more prominent for firms have different dialects from their provincial capital and a better judicial environment. provide additional evidence access to capital, innovation efficiency, city...