A5100451632- Robotics and Sensor-Based Localization
- Advanced Vision and Imaging
- Infrared Target Detection Methodologies
- Digital Media and Visual Art
- Advanced Image and Video Retrieval Techniques
- Photonic and Optical Devices
- Advanced Neural Network Applications
- 3D Surveying and Cultural Heritage
- Advanced Measurement and Detection Methods
- Optical measurement and interference techniques
- Image Enhancement Techniques
- Video Surveillance and Tracking Methods
- Remote Sensing and LiDAR Applications
- Atmospheric chemistry and aerosols
- Hydraulic Fracturing and Reservoir Analysis
- Semiconductor Lasers and Optical Devices
- Color perception and design
- Robotic Path Planning Algorithms
- Infrared Thermography in Medicine
- Remote Sensing and Land Use
- Cultural Heritage Materials Analysis
- Aerogels and thermal insulation
- Advanced Cellulose Research Studies
- Generative Adversarial Networks and Image Synthesis
- Domain Adaptation and Few-Shot Learning
PLA Information Engineering University
2022-2025
PLA Army Engineering University
2022-2025
Anhui Academy of Agricultural Sciences
2025
Taiyuan University of Technology
2022-2024
Xi'an University of Architecture and Technology
2023-2024
Yunnan University
2015-2024
Wuhan Ship Development & Design Institute
2024
Sanya University
2021-2024
Shenyang Jianzhu University
2022-2024
Duke University
2024
Efficiently and accurately separating infrared (IR) small targets from complex backgrounds presents a significant challenge. Numerous studies in the literature have proposed various feature fusion modules designed specifically to enhance extraction of IR target features. While these designs offer some incremental improvement accuracy detection, they come at steep cost significantly increasing network parameters FLOPs. Striving for balance between computational efficiency model accuracy, we...
To solve the problem of infrared (IR) small target tracking loss or error caused by factors such as scale changes, motion blur, occlusion, etc., this paper proposes a multi-strategy fusion algorithm using an IR segmentation network detection head, which mainly includes six strategies: pixel clustering, feature threshold adjustment, large area search, tracking, gate and coordinate solution. First, candidate targets are obtained through clustering strategy. Second, range is further reduced...
Effective detection of small targets plays a pivotal role in infrared (IR) search and track applications for modern military defense or attack. However, IR are very difficult to detect because their weak brightness, size, lack shape, structure, texture, other information elements. In order simultaneously satisfy the robustness timeliness target detection, inspired by density peak clustering human visual system, an idea combining improved global local contrast calculation is proposed. First,...
Abstract Ancient murals are important cultural heritages for our exploration of ancient civilizations and great research value. Due to long-time exposure the environment, often suffer from damage (deterioration) such as cracks, scratches, corrosion, paint loss, even large-region falling off. It is an urgent work protect restore these damaged murals. Mural inpainting techniques refer virtually filling deteriorated regions by reconstructing structure texture elements mural images. Most...
This paper presents the research results of detecting gastric polyps with deep learning object detection method in gastroscopic images. Gastric have various sizes. The difficulty polyp is that small are difficult to detect from background. We propose a feature extraction and fusion module combine it YOLOv3 network form our network. performs better than other methods because can fuse semantic information high-level maps low-level help detection. In this work, we use dataset created by...
Automatic urban area detection in remote sensing images is an important application the field of earth observation. Most existing methods employ feature classifiers and thereby contain a data training process. Moreover, some cannot detect areas complex scenes accurately. This letter proposes automatic method that uses multiple features have different resolutions. First, downsampled low-resolution image used to segment candidate area. After corner points are extracted, weighted Gaussian...
Robust segmentation in adverse weather conditions is crucial for autonomous driving. However, these scenes struggle with recognition and make annotations expensive, resulting poor performance. As a result, the Segment Anything Model (SAM) was recently proposed to finely segment spatial structure of provide powerful prior information, thus showing great promise resolving problems. SAM cannot be applied directly different geographic scales non-semantic outputs. To address issues, we propose...
Unmanned aerial vehicle (UAV) image localization can be used in global navigation satellite system (GNSS)-denied environments for UAV self-navigation. This study proposes a fast method images GNSS-denied based on sequence relationships. First, the LightGlue network extracts features from adjacent and combines these with relationship of to perform feature matching within overlapping range. After eliminating errors, affine transformation matrix is calculated realize positioning UAV....
Solid-state quantum emitters are pivotal for modern photonic technology, yet their inherent spectral inhomogeneity imposes a critical challenge in pursuing scalable network. Here, we develop cryogenic-compatible strain-engineering platform based on polydimethylsiloxane (PDMS) stamp that is not obviously working properly at cryogenic temperature. In-situ three-dimensional (3D) strain control achieved dots (QDs) embedded nanostructures. The compliant PDMS enables independent tuning of emission...
This paper examines the application of Long Short Term Memory (LSTM) model in music genre classification. We explore two different approaches paper. (1) In first method, we use one single LSTM to directly classify 6 genres music. The method is implemented and results are shown discussed. (2) approach only good for or less genres. So second approach, adopt a hierarchical divide- and-conquer strategy achieve 10 this classified into strong mild classes. Strong includes hiphop, metal, pop, rock...
As metaverse becomes one of the most popular buzzwords in technology, there is still a lack support to integrate true learning experiences massive open online courses (MOOCs). This article introduces new framework (MOMCs) and its major enabling technologies, which add immersive 3-D lacking MOOCs. It then describes detailed case study, President's First Lecture at Hong Kong University Science Technology (Guangzhou), we consider world's first MOMC environment, enabled by latest volumetric...