A5101855666- Human Mobility and Location-Based Analysis
- Recommender Systems and Techniques
- Advanced Vision and Imaging
- Digital Media Forensic Detection
- Advanced Measurement and Detection Methods
- Advanced Steganography and Watermarking Techniques
- Advanced Data Compression Techniques
- Network Security and Intrusion Detection
- Ionosphere and magnetosphere dynamics
- Remote Sensing and Land Use
- Infrared Target Detection Methodologies
- Robotics and Sensor-Based Localization
- Chaos-based Image/Signal Encryption
- Digital Media and Visual Art
- Anomaly Detection Techniques and Applications
- Color Science and Applications
- Advanced Image and Video Retrieval Techniques
- Magnetic confinement fusion research
- Color perception and design
- Particle accelerators and beam dynamics
- Cultural Heritage Materials Analysis
- Video Analysis and Summarization
- Spectroscopy and Chemometric Analyses
- Image and Signal Denoising Methods
- Image and Video Quality Assessment
University of Electronic Science and Technology of China
2021
Yunnan Normal University
2018-2020
University of Science and Technology of China
2017-2019
Tianjin University of Technology
2016-2018
Anhui University
2017
North China Electric Power University
2015
This paper designs a novel underpainting based camera shooting resilient (CSR) document watermarking algorithm for dealing with the leak source tracking problem. By applying such algorithm, we can extract authentication watermark information from candid photographs. The watermarked contains three significant properties. 1) Inconspicuousness. is inconspicuous and it will not easily be maliciously attacked. 2) Robustness. We propose DCT-based embedding distortion compensation extracting which...
Feature selection is one of the important factors that affect intrusion detection system.Aiming at problems due to selecting high feature dimension and redundancy causelow accuracy missing rate in traditional system.In this paper, deep belief network algorithm given select featureslayer by layer reduce dimension.As an unsupervised learning algorithm, it more suitable for features from a large number unlabeled data.Compared with other algorithm,the experiment shows effective than network.
A novel real time magnetic island identification system for HL-2A is introduced. The method based on the measurement of Mirnov probes and equilibrium flux constructed by fit (EFIT) code. consists an analog front board a digital processing connected shield cable. Four octal-channel analog-to-digital convertors are utilized 100 KHz simultaneous sampling all probes, applications PCI extensions Instrumentation platform reflective memory allow to receive EFIT results simultaneously. high...
Amphibious spherical robots that achieve the correct navigation need to collect surround information and clear perception. In addition equip various sensors, robot collecting can also rely on video images, which include rich texture of color. this paper, we present a image mosaic method surrounding amphibious robot. We extract key frame from based K-means clustering, control number by changing threshold. Meanwhile, pre-processing extracting points SIFT (Scale-invariant Feature Transform)....
An algorithm for signal extraction from a contaminated and distorted spectrum is proposed. First, this combines the salient space of statistical characteristics noise to detect regions at different scales. Second, it extracts signals by subtracting baseline in regions. The fitted segmented polynomial functions. This has been applied simulated experimental data, results show that can accurately automatically extract with varying widths spectrum. method minimizes influence distortion exhibits...
The principal component analysis method (PCA) and the kernel entropy (KECA) are used to construct spectral reflectance, study color reproduction. . This compares reconstruction precision through reflectance methods based on (PCA), (KPCA), (KECA). Experimental results show that algorithm KECA is superior than PCA KPCA in chromaticity precision. It has certain application value for true reproduction of object surface.
In recent years, deep neural network is introduced in recommender systems to solve the collaborative filtering problem, which has achieved immense success on computer vision, speech recognition and natural language processing. On one hand, can be used model auxiliary information systems. other it also capable of modeling nonlinear relationships between users items. One advantage that performance algorithm easily enhanced by augmenting depth network. However, two potential problems may emerge...