A5101491069- Video Surveillance and Tracking Methods
- Image Enhancement Techniques
- Advanced Steganography and Watermarking Techniques
- Chaos-based Image/Signal Encryption
- Robotics and Sensor-Based Localization
- Digital Media Forensic Detection
- UAV Applications and Optimization
- Energy, Environment, Economic Growth
- Electromagnetic Compatibility and Noise Suppression
- Industrial Vision Systems and Defect Detection
- Microbial Metabolic Engineering and Bioproduction
- Multimodal Machine Learning Applications
- Fire Detection and Safety Systems
- Advanced Neural Network Applications
- Robotic Path Planning Algorithms
- Advanced Manufacturing and Logistics Optimization
- Machine Learning in Bioinformatics
- Constraint Satisfaction and Optimization
- Visual Attention and Saliency Detection
- Evaluation and Optimization Models
- Precipitation Measurement and Analysis
- Optimization and Packing Problems
- Smart Parking Systems Research
- Blind Source Separation Techniques
- Leadership, Courage, and Heroism Studies
Jiangxi Normal University
2025
Jiangnan University
2023-2025
Zhejiang Lab
2024
Chinese Academy of Medical Sciences & Peking Union Medical College
2024
State Key Laboratory of Information Engineering in Surveying Mapping and Remote Sensing
2023
Wuhan University
2023
National University of Defense Technology
2019-2023
Xidian University
2019-2022
Wuhan University of Science and Technology
2017-2022
Minzu University of China
2011-2022
The Vision Meets Drone (VisDrone2019) Single Object Tracking challenge is the second annual research activity focusing on evaluating single-object tracking algorithms drones, held in conjunction with International Conference Computer (ICCV 2019). VisDrone-SOT2019 Challenge goes beyond its VisDrone-SOT2018 predecessor by introducing 25 more challenging sequences for long-term tracking. We evaluate and discuss results of 22 participating 19 state-of-the-art trackers collected dataset. are...
Remote sensing object tracking is a novel and challenging problem due to the negative effects of weak features background noise. In this paper, from perspective attention-focus deep learning, we propose Joint Siamese Attention-Aware Network (JSANet) for efficient remote which contains both self-attention cross-attention modules. First, modules emphasize interdependent channel-wise coefficient via channel attention conduct corresponding space transformation spatial domain information with...
BACKGROUND: Simple blood tests that could be used for early detection are crucial the ultimate control and prevention of colorectal cancer (CRC). In this study, we performed a serum proteomic analysis CRC health volunteers to identify novel biomarkers involved in CRC. METHOD: A shotg un method was applied proteins samples three using combination high-performance liquid chromatography mass spectrometry. Label-free protein profiling conducted quantify compare profiles volunteers. Two...
In this paper, a new reversible data hiding based on logarithm transformation function named improved reduced difference expansion (IRDE) was proposed. Reduced (RDE) proposed by Lou et al uses to reduce the distortion cased (DE) effectively. Inspired RDE, we modify embedding procedure and extraction correspondingly. The embedded can be extracted correctly, original image restored completely from marked images without distortion. Performance comparisons with Lou's scheme demonstrate...
In this article, a method for solving cable layout optimization problems in electronic systems based on ensemble learning (EL) and improved differential evolution (DE) algorithm is proposed. Using method, we produced two schemes with the considerations of electromagnetic compatibility performance total length cables. First, several different individual machine (ML) models all output variables were established. On basis, established EL better than previous ML models. Then, aiming at problem...
Robust estimation of frequency hopping radar signal parameters is the key to achieving stable jam in communication. In order deal with poor noise resistance performance conventional time-frequency analysis based parameter method under low signal-noise-ratio conditions. A robust proposed this paper for waveform estimation. The performed time domain. Firstly, pulse repetition interval received extracted by matching filtering, which used as window length divide into multiple frames. coherent...
In this paper, a novel reversible data hiding algorithm based on the prediction error is proposed. The different of host images calculated from direction according to embedded capacity. Multilayer technique used increase first layer horizontal scanning; second vertical scanning, and so on. Then predicted method predict target different. by changing error. Experimental results show that our scheme not only can embed easily, but also get good image's quality.
Risky driving behavior seriously affects the driver's ability to react, execute and judge, which is one of major causes traffic accidents. The timely accurate identification status drivers particularly important, since can quickly adjust their avoid safety In order further improve accuracy, this paper proposes a risky-driving image-recognition system based on visual attention mechanism deep-learning technology identify four types images including normal driving, while smoking, drinking...
Object-goal navigation is a challenging task that requires guiding an agent to specific objects based on first-person visual observations. The ability of comprehend its surroundings plays crucial role in achieving successful object finding. However, existing knowledge-graph-based navigators often rely discrete categorical one-hot vectors and vote counting strategy construct graph representation the scenes, which results misalignment with images. To provide more accurate coherent scene...
In male mammals, many developmental-stage-specific RNA transcripts (both coding and noncoding) are preferentially or exclusively expressed in the testis, where they play important roles spermatogenesis fertility. However, a reliable platform for efficiently depleting various types of to study their biological functions during
The traditional weld seam detection methods in passive vision are usually realized by detecting edges or textures of the seam. Since illumination conditions, types and backgrounds vary with tasks, these only work specific environment. This paper raises a method that replaces four main steps (pretreatment, thresholding, detection, fitting) an end-to end neural network system. eliminates ambiguity original horizontal bounding box adding inclination parameter to region proposal (RPN). Compared...
Style transfer has been widely used and developed in cultural entertainment industry. However, the traditional neural style cannot perfectly color of artistic images. To address this critical issue, we propose a color-first realtime framework. On one hand, ensure adequate transmission by giving priority to transfer; other design perceptual loss reduce style. Qualitative quantitative experiments demonstrate that compared with state-of-the-art methods, our framework can effectively images...
In ground-based observations of the Sun, solar images are often affected by appearance thin clouds, which contaminate and affect scientific results from data analysis. this paper, improved Pixel to Network (Pix2Pix) network is used convert polluted clear remove cloud shadow in images. By adding attention module model, hidden layer Pix2Pix model can infer map input feature vector according vector. And then, multiplied give different weights features map, adaptively refine make pay important...
Digital watermarking is a technique that embeds important data into host multimedia, such as video, image, audio, and so on, it can be used to protect the copyright of digital cover information. One its techniques reversible visible algorithm. It's not only authenticate reality information, but also recover original in some special images, medical military on. In this paper, we will summarize existing typical algorithms, give basic procedure them, analyze performance these algorithms.
There are a large number of cables in complex electronic systems, and these usually arranged multiple cable bundles with branches. In the layout bundles, total length weight all must be minimized, as much possible; addition, electromagnetic compatibility (EMC) performance system good possible. response to this problem, study proposes method suitable for branches systems: first, an automatic is proposed obtain initial bundles; then, four different machine learning methods used establish...
The Green Technology Innovation (GTI) is an effective way to deal with the problems of ecological environment. Taking New Energy Vehicle (NEV) technology as example, it has been explored that roles government, enterprises, public, universities and research institutions in green innovation by use big data. results have shown enterprise capacity was improved increasingly due large market demand more scientific participated field because policy support enlarged constantly recent years. However,...