Min Zhi

ORCID: 0009-0001-3069-8628
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Video Analysis and Summarization
  • Advanced Neural Network Applications
  • Human Pose and Action Recognition
  • Face and Expression Recognition
  • Advanced Image and Video Retrieval Techniques
  • Video Surveillance and Tracking Methods
  • Face recognition and analysis
  • Cooperative Communication and Network Coding
  • Visual Attention and Saliency Detection
  • Sports Analytics and Performance
  • Advanced MIMO Systems Optimization
  • Music and Audio Processing
  • Image Retrieval and Classification Techniques
  • Handwritten Text Recognition Techniques
  • Anomaly Detection Techniques and Applications
  • Vehicle License Plate Recognition
  • Telecommunications and Broadcasting Technologies
  • Multimedia Communication and Technology
  • Sports Dynamics and Biomechanics
  • Biometric Identification and Security
  • Video Coding and Compression Technologies
  • Gait Recognition and Analysis
  • Advanced Wireless Network Optimization
  • Industrial Vision Systems and Defect Detection
  • Advanced Mathematical Theories

Inner Mongolia Normal University
2009-2024

Inner Mongolia University
2021

China Coal Research Institute (China)
2021

Central China Normal University
2016-2017

Pingjin Hospital
2012

Yunnan Vocational College of Mechanical and Electrical Technology
2012

Keio University
2009

Beijing University of Posts and Telecommunications
2005

Scene text detection is a fundamental research work in the field of image processing and has extensive application value. Segmentation-based methods have time-consuming feature processing, while post-processing algorithms are excellent. Real-time semantic segmentation use lightweight backbone networks for extraction aggregation but lack effective methods. The pure convolutional network improves model performance by changing key components. Combining advantages three types methods, we propose...

10.3390/electronics12143055 article EN Electronics 2023-07-12

In multiuser MIMO-BC (Multiple-Input Multiple-Output Broadcasting) systems, user selection is important to achieve diversity. The optimal algorithm try all the combinations of users find group that can However, calculation amount too large implement. Thus, instead algorithm, some suboptimal algorithms were proposed based on semiorthogonality channel vectors. purpose this paper diversity with a small calculation. For purpose, we propose improve orthogonality selected group. Simulation results...

10.1109/pimrc.2009.5449749 article EN 2009-09-01

Blended learning has caused comprehensive attention for domestic and foreign researchers with the rapid development of information technology theory lifelong learning. In view inevitable shortcomings traditional offline teaching e-learning, an exploration blended pattern based on Hstar platform smart classroom is proposed which a new education idea. this paper, we select software engineering course use as supporting environment, where students are allowed to choose content according their...

10.1109/iset.2017.10 article EN 2017-06-01

Object detection is one of the most basic and central task in computer vision. Its to find all interested objects image, determine category location objects. widely used has strong practical value research prospects. Applications include face detection, pedestrian vehicle detection. In recent years, with development convolutional neural network, significant breakthroughs have been made object This paper describes detail classification algorithms based on deep learning. The are mainly divided...

10.1117/12.2557219 article EN 2020-01-03

Video summary technology has become a hotspot of current researches. The application sports video can quickly fetch important information in that help enthusiasts and senior analysis the video. present study takes tennis as research object. Firstly, determine number key frames based on statistical rules, then extract from different kinds shots. Secondly, give coefficient to through function audio middle-level features. Finally, divide into levels according generate forms. generation which...

10.1109/icsess.2015.7339134 article EN 2015-09-01

With the gradual improvement of level educational information, idea education big data has gone deep into people's minds. More and more teaching staff researchers have consciousness needs, expect to do mining learning analysis data. However, most realistic basic problem they in process carrying out research is that there no good support or a convenient way get Aiming at problem, this article puts forward an open service model, builds platform based on model. The tried run Central China...

10.1109/icccbda.2017.7951903 article EN 2017-04-01

Attention mechanism is one of the most basic and core tasks in computer vision. Its essence to locate information region interest suppress useless information. The results are usually displayed form probability graph or eigenvector. has become an important concept convolutional neural network, which been widely studied different application fields strong practical value. This paper introduces classification attention its fine-grained image recognition. mainly divided into channel mechanism,...

10.1117/12.2623383 article EN Thirteenth International Conference on Graphics and Image Processing (ICGIP 2021) 2022-02-16

Shot change detection is one of the critical techniques in video browsing and indexing system. In this paper, we propose a shot algorithm with adaptive thresholds on DC images. The bit-rate information used to decrease number I frames that take part detection. Then, local are obtained based main color segment sequence. Finally, changes detected modified twin-comparison method. Experimental results show our can detect from long sequence reduced computation complexity, at same time solve...

10.1109/iwvdvt.2005.1504572 article EN 2005-09-09

Key frames are representative in a shot. With key frames, video information can be efficiently managed retrieval and indexing. In view of the scenery characteristic user attention focus, this paper proposed new semantic based frame extraction method. First according to amplitude angle P frame's motion vector, shot is judged; then speed camera motion, duration distribution extracted. The experiment result shows that reflect content well.

10.1109/icwapr.2007.4420729 article EN International Conference on Wavelet Analysis and Pattern Recognition 2007-01-01

In multiuser MIMO-BC (Multiple-Input Multiple-Output Broadcasting) systems, user selection is important to achieve diversity. The optimal algorithm try all the combinations of users find group that can Unfortunately, high calculation cost prevents its implementation. Thus, instead algorithm, some suboptimal algorithms were proposed based on semiorthogonality channel vectors. purpose this paper diversity with a small amount calculation. For purpose, we propose improve orthogonality selected...

10.1587/transcom.e92.b.2667 article EN IEICE Transactions on Communications 2009-01-01

In response to the slow running speed of Deformable Part Model algorithm in process pedestrian detection, this paper incorporated Cascade Detection and Branch-and-Bound into a fast detection which is based on Model. process, sequence model evaluates individual parts sequentially quickly prune most smaller possible objects. This aims accelerate object positioning, optimize global classification results all image regions. Meanwhile, boundaries maximum are adopted search clipping operation...

10.1117/12.2281594 article EN Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE 2017-07-21

The construction of deep neural networks depends on a significant number parameters and computational complexity, which poses challenge in the field image processing. To address issue Transformer network model's large size inability to effectively capture local features image, this paper proposes lightweight composite structure that combines spectral feature refinement module (SFRM) parameterless attention augmentation (PAAM). SFRM PAAM work together improve quality used transformer....

10.1145/3652583.3658006 article EN 2024-05-30

10.1109/ijcnn60899.2024.10651133 article EN 2022 International Joint Conference on Neural Networks (IJCNN) 2024-06-30

10.1109/ijcnn60899.2024.10650145 article EN 2022 International Joint Conference on Neural Networks (IJCNN) 2024-06-30

The purpose of cross-age face recognition is to identify people with a large age difference. It great significance in security, finance, fighting criminals and other fields. In order organize the development process Convolutional Neural Networks (CNN) field recognition, main discriminant methods recent years are summarized. First all, evolution performance based on CNN systematically described, Secondly, advantages disadvantages most commonly used data sets introduced. Finally, combined...

10.1109/icivc58118.2023.10270506 article EN 2022 7th International Conference on Image, Vision and Computing (ICIVC) 2023-07-27

In this paper SVM algorithm is applied to classify the scenery video types in compressed domain. Firstly we extract sequences randomly from and detect representative frames sequences; secondly features such as color layout, dominant color, edge histogram face feature; then according SVM, are classified natural scenery, personality, animal plant. Experimental results have shown that result of our very high accuracy.

10.1109/primeasia.2009.5397389 article EN 2009-11-01

In order to solve the existing problems of noise-sensitive and inefficient background updating in current object detection process, an effective moving method which based on two layers model is proposed this paper. The codebook using brightness color features presented first layer, while a running average algorithm second we get target by comprehensive weighted according layers' results. Experimental results demonstrate that can detect more accurately, completely, update faster, it shows...

10.1109/icinis.2015.49 article EN 2015-11-01
Coming Soon ...