A5102006880- Image Retrieval and Classification Techniques
- Ferroelectric and Piezoelectric Materials
- Multiferroics and related materials
- Medical Image Segmentation Techniques
- Remote-Sensing Image Classification
- Image Processing and 3D Reconstruction
- Video Surveillance and Tracking Methods
- Dielectric properties of ceramics
- Advanced Image Fusion Techniques
- Face and Expression Recognition
- Advanced Image and Video Retrieval Techniques
- Human Pose and Action Recognition
- Advanced Neural Network Applications
- Microwave Dielectric Ceramics Synthesis
- Generative Adversarial Networks and Image Synthesis
- Image and Object Detection Techniques
- Advanced Vision and Imaging
- Dielectric materials and actuators
- Handwritten Text Recognition Techniques
- Maritime Navigation and Safety
- Image and Signal Denoising Methods
- Thermal Radiation and Cooling Technologies
- Multimodal Machine Learning Applications
- Advanced Graph Neural Networks
- Infrared Target Detection Methodologies
Harbin Engineering University
2014-2024
China Three Gorges University
2022-2024
China Energy Engineering Corporation (China)
2024
Ministry of Industry and Information Technology
2023-2024
Harbin Institute of Technology
2006-2023
Harbin University
2011-2019
State Council of the People's Republic of China
2019
Harbin University of Commerce
2017
Harbin University of Science and Technology
2013
Daqing Oilfield General Hospital
2010
In this study, we present an effective strategy to enhance energy-storage density by the Mn2+ substitution of Ti4+ into 0.7(Na0.5Bi0.5)TiO3-0.3SrTiO3 (0.7NBT-0.3ST) relaxor ferroelectric thin films. The influence Mn doping on microstructures, properties, and performances as-prepared films was investigated in detail. results show that values electric break-down field strength difference between maximum polarization remnant are markedly improved appropriate doping. Owing high 1894 kV/cm huge...
Federated learning can effectively ensure data security and improve the problem of islanding. However, performance federated learning-based schemes could be better due to imbalance image data. Therefore, this paper proposes a approach based on priori knowledge bilateral segmentation network for edge extraction. First, distribute training images some special complex small sample unshared Then, with similar information original is learned obtain prior knowledge, local uniform sparsity method...
Srx(Na0.5Bi0.5)1−xTi0.99Mn0.01O3 (x = 0.2, 0.4, 0.6, and 0.8) relaxor ferroelectric thin films were grown on Pt/Ti/SiO2/Si substrates by the Sol-Gel method. The influence of Sr content microstructures, properties, energy-storage performances investigated in detail. Sr0.6(Na0.5Bi0.5)0.4Ti0.99Mn0.01O3 film exhibits very slim hysteresis loops with highest electric breakdown field strength due to reduced oxygen vacancies. Owing high 3134.3 kV/cm, shows a giant recoverable density 33.58 J/cm3....
High energy storage density and a reversible electrocaloric effect are simultaneously achieved in Sr<sub>0.995</sub>(Na<sub>0.5</sub>Bi<sub>0.5</sub>)<sub>0.005</sub>(Ti<sub>0.99</sub>Mn<sub>0.01</sub>)O<sub>3</sub> amorphous thin films <italic>via</italic> polar cluster engineering.
Dielectric capacitors have received more and attention because of fast charge/discharge capability. However, the energy-storage performance still cannot meet demand. In this work, lead-free perovskite Sr1-x(Na0.5Bi0.5)xTi0.99Mn0.01O3 (x = 0, 0.005, 0.01, 0.02) thin films prepared by sol-gel method were carefully studied. Defect dipoles local lattice distortion created doping Mn at B-site, enabling ferroelectric polarization behavior. To further enhance polarization, co-substitution A-site...
Capacitors with high energy storage density, low cost, ultrafast charge–discharge capability, and environmental friendliness are in demand for application new vehicles, modern electrical systems, high-energy laser weapons.
The edge-based active contour model has been one of the most influential models in image segmentation, which level set method is usually used to minimize energy function and then find desired contour. However, for infrared thermal pedestrian images, traditional set-based that utilizes gradient information as edge indicator fails provide satisfactory boundary target. That due poorly defined boundaries intensity inhomogeneity. Therefore, we propose a novel segmentation able deal with above...
In this study, an effective method is presented to enhance the dielectric constant at high temperature through Ba2+, Mg2+ codoped nonstoichiometric Bi0.5Na0.5TiO3 (BNT) relaxor ferroelectric ceramics. The dependence of permittivity, complex impedance spectrum, and conductivity was investigated understand effect doping on properties BNT. results show that value (Ba0.01Bi0.5Na0.49)(Ti0.9Mg0.1)O3 ceramics increases 2 orders magnitude with a very broad range corresponding pure BNT spectrum...
Traditional methods for estimating optical flow use variational model that includes data term and smoothness term, which can build a constraint relationship between two adjacent images flow. However, most of them are too slow to be used in real-time applications. Recently, convolutional neural networks have been area successfully. Many current learning large sets contain ground truth network training, make prior knowledge estimate directly. these overemphasize the factor deep ignore...
Passive daytime radiative cooling (PDRC) technique can cool objects without energy consumption by reflecting sunlight and radiating heat into the cold outer space through atmospheric transparent windows (ATW). Here, 3D porous polydimethylsiloxane (P-PDMS) films with excellent properties were fabricated using citric acid monohydrate (CAM) templates. Integrating high reflectance (78.6%) great thermal emissivity (99.4%) in ATW, test was performed an outdoor environment average solar radiation...
The automatic text detection in document images is useful for many applications. This paper presents an algorithm that can automatically detect and extract digital images. Firstly, we process fuse Gabor filtered at different orientations scales obtain image reflects the layout of image. Then, potential regions are directly extracted from resulting Finally, two criteria based on geometrical property high frequency content adopted to kick-out those non-text regions. experiments performed some...
The automatic text detection in document images is useful for many applications. This paper presents an algorithm that can automatically detect and extract digital images. Firstly, we process fuse Gabor filtered at different orientations scales obtain image reflects the layout of image. Then, potential regions are directly extracted from resulting Finally, two criteria based on geometrical property high frequency content adopted to kick-out those non-text regions. experiments performed some...
Moving object detection means segmenting objects that exhibit motion in image sequences. Several deep neural network-based methods have been proposed and almost perfect performance. However, when these algorithms are used on unseen videos, their performance drops dramatically. In this paper, we propose a moving method based Graph Neural Network(GNN) for videos called IGMOD. Specifically, use Mask-RCNN instances, the temporal median filter to initialize background extract optical flow,...