A5101861780- Advanced Image and Video Retrieval Techniques
- Robotics and Sensor-Based Localization
- Medical Imaging Techniques and Applications
- Advanced MRI Techniques and Applications
- Image and Object Detection Techniques
- Advanced Vision and Imaging
- Biometric Identification and Security
- Radiomics and Machine Learning in Medical Imaging
- Plasmonic and Surface Plasmon Research
- Metamaterials and Metasurfaces Applications
- Remote-Sensing Image Classification
- Advanced biosensing and bioanalysis techniques
- Advanced X-ray Imaging Techniques
- Electron and X-Ray Spectroscopy Techniques
- Photonic and Optical Devices
- Advanced Image Fusion Techniques
- Terahertz technology and applications
- Blockchain Technology Applications and Security
- Orbital Angular Momentum in Optics
- Satellite Image Processing and Photogrammetry
- IoT and Edge/Fog Computing
- Mathematical Approximation and Integration
- Advanced Image Processing Techniques
- Cybersecurity and Cyber Warfare Studies
- Superconducting and THz Device Technology
PLA Information Engineering University
2020-2024
Zaozhuang University
2024
Beihang University
1988-2018
Wake Forest University
2012-2016
Nanjing University of Posts and Telecommunications
2012
Syracuse University
2007-2010
Embedded and mobile smart devices face problems related to limited computing power excessive consumption. To address these problems, we propose Mixed YOLOv3-LITE, a lightweight real-time object detection network that can be used with non-graphics processing unit (GPU) devices. Based on YOLO-LITE as the backbone network, YOLOv3-LITE supplements residual block (ResBlocks) parallel high-to-low resolution subnetworks, fully utilizes shallow characteristics while increasing depth, uses "shallow...
We demonstrate a three-axis atomic magnetometer with one intensity-modulated pump beam and orthogonal probe beam. The main field component is measured using the resonance of pumping light, while transverse components are simultaneously optical rotation modulated by spin precession. It an all-optical without any modulation or radio frequency field. Magnetic sensitivity 0.8 pT/Hz1∕2 achieved under bias 2 μT.
The quasi-bound states in the continuum (quasi-BIC) are uniquely attractive fields of nonlinear modulators, optical switches, and sensing due to their ultra-high radiative quality factor. implementation of...
We consider the problem of matching highly non-ideal ocular images where iris information cannot be reliably used. Such are characterized by non-uniform illumination, motion and de-focus blur, off-axis gaze, non-linear deformations. To handle these variations, a single feature extraction scheme is not sufficient. Therefore, we propose an fusion framework three distinct schemes utilized in order to significant variability input images. The Gradient Orientation Histogram (GOH) extracts global...
We present a randomized singular value decomposition (rSVD) method for the purposes of lossless compression, reconstruction, classification, and target detection with hyperspectral (HSI) data. Recent work in low‐rank matrix approximations obtained from random projections suggests that these are well suited dimensionality reduction. Approximation errors rSVD evaluated on HSI, comparisons made to deterministic techniques as other approximation methods involving compressive principal component...
A hybrid thrust magnetic bearing (HTMB) with permanent magnet creating bias flux is used to save energy. The eddy currents have a fundamental impact on the dynamic characteristics of system. Therefore, factor models affected by rather than static model are required insure reliable and effective operation. for HTMB magnets subsidiary air gap presented in this paper. equivalent circuit given derived calculate force-current force-displacement factor. Furthermore, factors also investigated...
Traditionally, the correction of panoramic images has been achieved through sundial projection or cubic projection. However, these methods have not adequately addressed issues overlap and fixed pose relationships in corrected images. To address this, we propose an enhanced algorithm for correcting using The transformation process involves mapping spherical image onto a cube that is tangent to sphere, with cube's side length exceeding diameter sphere. For re-projection experiments, selected...
technologies coupled with rapid growth in computing power
A digital twin city maps a virtual three-dimensional (3D) model to the geographic information system, constructs world, and integrates real sensor data achieve purpose of virtual–real fusion. Focusing on accuracy problem vision registration in scene, this study proposes 2D-3D method under inverse depth structural semantic constraints. First, perspective images scene were obtained by using view inverse-depth nascent technology, then features extracted two-line minimal solution set method....
Abstract Visual place recognition (VPR) involves obtaining robust image descriptors to cope with differences in camera viewpoints and drastic external environment changes. Utilizing multiscale features improves the robustness of descriptors; however, existing methods neither exploit generated during feature extraction nor consider redundancy problem when fusing information are enhanced. We propose a novel encoding strategy—convolutional multilayer perceptron orthogonal fusion...
Building instance extraction and recognition (BEAR) extracts further recognizes building instances in unmanned aerial vehicle (UAV) images, holds with paramount importance urban understanding applications. To address this challenge, we propose a unified network, BEAR-Former. Given the difficulty of due to small area multiple UAV developed novel multi-view learning method, Cross-Mixer. This method constructs cross-regional branch an intra-regional to, respectively, extract global context...
Social media data are characterized by significant noise and non-standardization, thereby posing challenges for existing methods in recognizing named entities owing to the entity sparsity insufficient semantic richness. Thus, deal with these issues, this study proposes SEMFF-NER, a recognition (NER) method social texts that integrates multi-scale features syntactic information. First, global extracted using Transformer-based encoder (XLNET) embedded dependency relations enhance...
Geolocating a street-view image by matching it with geotagged satellite images is crucial for location assessment. However, the perspective disparity between and presents significant challenges. To address issue, mainstream approach to convert ground-level perspective. So reference are not only required center but also coverage be consistent street view sometimes even consistency in north direction, which difficult achieve practical applications. This paper introduces ground-breaking method...
Cross-view geolocation, which aims to geolocate ground-view images using reference satellite imagery, is a challenging task that requires effective strategies minimize significant disparities between images. In this paper, we propose novel approach address challenge. By exploiting the projection relationship ground and images, are able convert into satellite-view thereby mitigating inherent two perspectives. To enhance converted further, employ conditional generative adversarial network...
Obtaining the geographical location of images through image geo-localization technology is a highly significant task. However, existing methods struggle with accuracy under difficult conditions such as viewpoint changes, illumination variations, seasonal and occlusions. To address these challenges, we proposed an architecture based on foundation vision model feature mixing. The involves truncating fine-tuning DINOv2 to extract robust features, which are then aggregated using MLP-Mixer-based...