A5100702368- Chaos-based Image/Signal Encryption
- Sparse and Compressive Sensing Techniques
- Advanced Image Processing Techniques
- Image and Signal Denoising Methods
- Image Enhancement Techniques
- Advanced Steganography and Watermarking Techniques
- Advanced Vision and Imaging
- Stochastic Gradient Optimization Techniques
- Cellular Automata and Applications
- Image Processing Techniques and Applications
- Robotic Path Planning Algorithms
- Manufacturing Process and Optimization
- Domain Adaptation and Few-Shot Learning
- Photoacoustic and Ultrasonic Imaging
- Chaos control and synchronization
- Artificial Immune Systems Applications
- Thermography and Photoacoustic Techniques
- Essential Oils and Antimicrobial Activity
- Cultural Heritage Materials Analysis
- Soil, Finite Element Methods
- Data Management and Algorithms
- Heat Transfer and Boiling Studies
- Underwater Vehicles and Communication Systems
- Remote-Sensing Image Classification
- Advanced Algorithms and Applications
Dalian University of Technology
2013-2021
City Institute, Dalian University of Technology
2017-2021
Sichuan Agricultural University
2014
South China University of Technology
2011
Chongqing Normal University
2003-2006
Southwest University
2006
Numerous tasks at the core of statistics, learning and vision areas are specific cases ill-posed inverse problems. Recently, learning-based (e.g., deep) iterative methods have been empirically shown to be useful for these Nevertheless, integrating learnable structures into iterations is still a laborious process, which can only guided by intuitions or empirical insights. Moreover, there lack rigorous analysis about convergence behaviors reimplemented iterations, thus significance such little...
Abstract In this paper, a novel image encryption algorithm based on the Once Forward Long Short Term Memory Structure (OF-LSTMS) and Two-Dimensional Coupled Map Lattice (2DCML) fractional-order chaotic system is proposed. The original divided into several blocks, each of which input OF-LSTMS as pixel sub-sequence. According to sequences generated by 2DCML system, parameters gate, output gate memory unit are initialized, positions changed at same time changing values, achieving...
We present a new data-driven method for robust skin detection from single human portrait image. Unlike previous methods, we incorporate body as weak semantic guidance into this task, considering acquiring large-scale of labeled data is commonly expensive and time-consuming. To be specific, propose dual-task neural network joint via semi-supervised learning strategy. The contains shared encoder but two decoders separately. For each decoder, its output also serves counterpart, making both...
Blind image deblurring plays a very important role in many vision and multimedia applications. Most existing works tend to introduce complex priors estimate the sharp structures for blur kernel estimation. However, it has been verified that directly optimizing these models is challenging easy fall into degenerate solutions. Although several experience-based heuristic inference strategies, including trained networks designed iterations, have developed, still hard obtain theoretically...
This paper presents a tabu insertion search algorithm (TIS) based on the merits of method (IM) and (TS) for solving travel salesman problem (TSP). TIS combines good local ability with tour construct to solution space solve combinatorial optimization problem. In this paper, we use classical NP hard TSP test effectiveness TIS. Results show that has jump beyond optimum obtain global optimum.
Path planning for mobile robots is an important topic in modern robotics. This paper proposes a novel approach to path problem robots, which the model of vertexes obstacles constructed describe two-dimensional map work place robot order obtain constrained function from start location goal location, and further translates into nonlinear minimum optimization problem, finally using modified particle swarm optimizer optimizes forcefully get satisfactory collision-free solution. The effectiveness...
Operator splitting methods have been successfully used in computational sciences, statistics, learning and vision areas to reduce complex problems into a series of simpler subproblems. However, prevalent schemes are mostly established only based on the mathematical properties some general optimization models. So it is laborious process often requires many iterations ideation validation obtain practical task-specific optimal solutions, especially for nonconvex real-world scenarios. To break...
Operator splitting methods have been successfully used in computational sciences, statistics, learning and vision areas to reduce complex problems into a series of simpler subproblems. However, prevalent schemes are mostly established only based on the mathematical properties some general optimization models. So it is laborious process often requires many iterations ideation validation obtain practical task-specific optimal solutions, especially for nonconvex real-world scenarios. To break...
The overall performance of Particle Swarm Optimizer lies on its ability to harmonize global and local search process. By dividing the whole swarm into equal sub-swarms with iterative cooperation, taking a series Sugeno functions as inertia weight decline curves for each sub-swarm, an adaptive strategy was proposed adaptively select different curve according vary rate sub-swarm's fitness value. Experimental results several benchmark show that modified algorithm can effectively balance avoid...
Numerous tasks at the core of statistics, learning and vision areas are specific cases ill-posed inverse problems. Recently, learning-based (e.g., deep) iterative methods have been empirically shown to be useful for these Nevertheless, integrating learnable structures into iterations is still a laborious process, which can only guided by intuitions or empirical insights. Moreover, there lack rigorous analysis about convergence behaviors reimplemented iterations, thus significance such little...
Based on MTS-810 type vibration testing machine, morphological characteristics of hysteretic curves frozen clay are quantitatively studied, and dynamic mechanical response analyzed consisting stiffness, viscosity, degree microscopic damage, residual strain energy dissipation. The studies have shown that the higher frequencies are, greater stiffness is, while smaller dissipation are. Stiffness, less affected by confining pressure. With increasing stress amplitude, decreases gradually,...
Distributed underwater sensor network coverage is divided into two main categories: deterministic and stochastic coverage. A strategy put forward to deploy determinate area by using a triangular-grid method. When ratio known, the distance between nodes can be adjusted meet in monitored area, least number of calculated. Also heuristic method proposed for deployment strategy. It under premise that initial node location randomly deployed given, Voronoi diagram, not easiest path searched,...
Abstract In this paper, we propose a new and efficient lightweight DenseNet which optimizes the parameter redundancy high FLOPs in model. According to distribution of weight value, element Lightweight Mix-Structure Convolution (LMSC) is realized model, reduces calculation required for model construction, ensures that accuracy does not decline significantly. The experimental results show compared with DenseNet-40-P only uses 45.5% parameters 54.3% FLOPs, reduced by less than 0.4%.
By analyzing and studying the most current algorithms about mining association rule, rules evaluated by minimum confidence could not ensure validity of will generate unrelated which affect intrusion detection work. This paper proposes CF measure based on previous work applies rule algorithm to technology detect behaviors in network. Finally, experiments show that improved is more efficient.
With mobile platform development there are more and Android-based image processing applications. The principles of four kinds edge detection algorithms analyzed in this paper such realized by adopting JNI technology based on android platform. At last the effect efficiency also compared summarized.
the paper suggests a kind of self-adapted layer contrast enhancement algorithm for medical images, which, in reference to Laplacian pyramid function, could compose and enhance pixel on each through F/E processing use self-adaptation Sigmoid it uses factor control degree. The result shows that details images can be displayed clearly with this without selecting characteristic center dimensions. It will not generate artifacts thus improves visual effect images.