A5066878094- Image Processing and 3D Reconstruction
- Advanced Neural Network Applications
- 3D Shape Modeling and Analysis
- Remote-Sensing Image Classification
- 3D Surveying and Cultural Heritage
- Remote Sensing and Land Use
- Metallurgy and Material Forming
- Meteorological Phenomena and Simulations
- Enhanced Oil Recovery Techniques
- Visual Attention and Saliency Detection
- Image and Signal Denoising Methods
- HVDC Systems and Fault Protection
- Hydrology and Watershed Management Studies
- Advanced Drug Delivery Systems
- biodegradable polymer synthesis and properties
- Pickering emulsions and particle stabilization
- High Temperature Alloys and Creep
- Microstructure and Mechanical Properties of Steels
- Arctic and Antarctic ice dynamics
- Hydrological Forecasting Using AI
- Surfactants and Colloidal Systems
- CCD and CMOS Imaging Sensors
- Nanoparticle-Based Drug Delivery
- Advanced Image Fusion Techniques
- High-Voltage Power Transmission Systems
University of Science and Technology Beijing
2023-2025
Beijing Institute of Technology
2024
Institute of New Materials
2024
Chongqing University of Technology
2024
Tsinghua University
2022
Dalian University of Technology
2005-2021
Polyacrylamide-based inverse emulsions are one class of fascinating products possessing excellent aqueous thickening ability, rapid solubility, and tailorable properties in many industrial fields, but further improvement their performances simplification applicational operations still challenging. Traditionally, post-addition an inverting agent into the produced emulsion is necessary to enable product become soluble water. In this study, cetylmethyldiallylammonium chloride (CMDAAC) stearyl...
Hyperspectral remote sensing images are widely used in resource exploration, urban planning, natural disaster assessment, and feature classification. Aiming at the problems of poor interpretability classification algorithms for hyperspectral images, multiple dimensions, difficulty effectively improving accuracy, this paper proposes a band adaptive selection method images. The proposed model focuses on joint salient regions image, visualizes contribution bands, more intuitively reveals basis...
Abstract. Accurate lead-time forecast of ice breakup is one the key aspects for flood prevention and reducing losses. In this paper, a new data-driven model based on Statistical Learning Theory was employed prediction. The model, known as Support Vector Machine (SVM), follows principle that aims at minimizing structural risk rather than empirical risk. order to estimate appropriate parameters SVM, Multiobjective Shuffled Complex Evolution Metropolis (MOSCEM-UA) algorithm performed through...
As an important structure in the convolutional neural network, pooling layer can effectively reduce generalization of improved features after convolution. However, current method is easy to lose key low-frequency small sample data for unbalanced hyperspectral image data, which leads a decrease classification accuracy such data. In this paper, we propose method, GL-pooling (global-local pooling), target category samples. Calculating joint probability samples global and local maximizing...
We revisit the relationship between attention mechanisms and large kernel ConvNets in visual transformers propose a new spatial named Large Kernel Convolutional Attention (LKCA). It simplifies operation by replacing it with single convolution. LKCA combines advantages of convolutional neural networks transformers, possessing receptive field, locality, parameter sharing. explained superiority from both convolution perspectives, providing equivalent code implementations for each view....
As part of human core knowledge, the representation objects is building block mental that supports high-level concepts and symbolic reasoning. While humans develop ability perceiving situated in 3D environments without supervision, models learn same set abilities with similar constraints faced by infants are lacking. Towards this end, we developed a novel network architecture simultaneously learns to 1) segment from discrete images, 2) infer their locations, 3) perceive depth, all while...
The regular spherical poly-D, L-lactic acid (PDLLA) nanoparticles containing DNR are produced using a double-emulsion solvent evaporation process. products morphology is characterized by laser light-scattering particle size analyzer and scanning electron microscopy. Results show mean diameter in 546.9 nm with narrow distribution homogeneous production. encapsulation efficiency evaluated UV spectra. results indicate that drug contents loading 14.34% 72.9% respectively, the release profile...
The protection thyristor is commonly applied in the high voltage direct current modular multilevel converter (MMC-HVDC) for handling with extreme shoot-through faults. To reduce building complexity and costs, a novel solution of MMC-HVDC system based on integrated gate commutating (IGCT) central overvoltage breakdown function proposed. Estimation fault equivalent model considering stray inductances shows that di/dt limiting inductor IGCT MMC sub-module (IGCT-SM) can help limit current....