Xiaoguang Lu

ORCID: 0000-0003-4702-589X
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Face recognition and analysis
  • Face and Expression Recognition
  • Medical Image Segmentation Techniques
  • Synthetic Aperture Radar (SAR) Applications and Techniques
  • Image and Signal Denoising Methods
  • Precipitation Measurement and Analysis
  • Advanced SAR Imaging Techniques
  • Biometric Identification and Security
  • Advanced Neural Network Applications
  • Meteorological Phenomena and Simulations
  • Radar Systems and Signal Processing
  • Microfluidic and Bio-sensing Technologies
  • Microfluidic and Capillary Electrophoresis Applications
  • Blind Source Separation Techniques
  • Advanced Numerical Methods in Computational Mathematics
  • Remote-Sensing Image Classification
  • Image Processing Techniques and Applications
  • Direction-of-Arrival Estimation Techniques
  • Medical Imaging and Analysis
  • Air Traffic Management and Optimization
  • Speech and Audio Processing
  • Soil Moisture and Remote Sensing
  • AI in cancer detection
  • Advanced Image Fusion Techniques
  • Medical Imaging Techniques and Applications

AiCure (United States)
2024

Civil Aviation University of China
2009-2024

Changchun University of Science and Technology
2024

Parthenope University of Naples
2023-2024

Wuhan University
2021-2023

Singapore University of Technology and Design
2019-2022

Changchun University
2019

China Academy of Railway Sciences
2017

Tongji Hospital
2017

Huazhong University of Science and Technology
2017

The performance of face recognition systems that use two-dimensional images depends on factors such as lighting and subject's pose. We are developing a system utilizes three-dimensional shape information to make the more robust arbitrary pose lighting. For each subject, 3D model is constructed by integrating several 2.5D scans which captured from different views. simplified (x,y,z) surface representation contains at most one depth value (z direction) for every point in (x, y) plane. Two...

10.1109/tpami.2006.15 article EN IEEE Transactions on Pattern Analysis and Machine Intelligence 2006-01-01

Face recognition based on 3D surface matching is promising for overcoming some of the limitations current 2D image-based face systems. The shape generally invariant to pose and lighting changes, but not non-rigid facial movement, such as expressions. Collecting storing multiple templates account various expressions each subject in a large database practical. We propose modeling scheme match 2.5D scans presence both deformations changes (multiview) template. A hierarchical geodesic-based...

10.1109/tpami.2007.70784 article EN IEEE Transactions on Pattern Analysis and Machine Intelligence 2008-06-20

The performance of face recognition systems that use two-dimensional (2D) images is dependent on consistent conditions such as lighting, pose and facial expression. We are developing a multi-view system utilizes three-dimensional (3D) information about the to make more robust these variations. This work describes procedure for constructing database 3D models matching this 2.5D scans which captured from different views, using coordinate invariant properties surface. simplified (x, y, z)...

10.1109/icpr.2004.915 article EN Deleted Journal 2004-08-23

Human facial images provide the demographic information, such as ethnicity and gender. Conversely, gender also play an important role in face-related applications. Image-based identification problem is addressed a machine learning framework. The Linear Discriminant Analysis (LDA) based scheme presented for two-class (Asian vs. non-Asian) classification task. Multiscale analysis applied to input images. An ensemble framework, which integrates LDA face at different scales, proposed further...

10.1117/12.542847 article EN Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE 2004-08-25

Current two-dimensional face recognition approaches can obtain a good performance only under constrained environments. However, in the real applications, appearance changes significantly due to different illumination, pose, and expression. Face recognizers based on representations of input images have sensitivity these variations. Therefore, combination classifiers which integrate complementary information should lead improved classification accuracy. We use sum rule RBF-based integration...

10.1109/icme.2003.1221236 article EN 2003-01-01

Current 2D face recognition systems encounter difficulties in recognizing faces with large pose variations. Utilizing the pose-invariant features of 3D data has potential to handle multiview matching. A feature extractor based on directional maximum is proposed estimate nose tip location and angle simultaneously. profile model represented by subspaces used select best candidates for tip. Assisted a statistical model, multimodal scheme presented extract eye mouth corners. Using automatic...

10.1109/fgr.2006.23 article EN 2006-04-28

An independent component analysis (ICA) based approach is presented for learning view-specific subspace representations of the face object from multiview examples. ICA, its variants, namely (ISA) and topographic (TICA), take into account higher order statistics needed view characterization. In contrast, principal (PCA), which de-correlates second moments, can hardly reveal good features characterizing different views, when training data comprises a mixture examples done in an unsupervised...

10.1109/tip.2005.847295 article EN IEEE Transactions on Image Processing 2005-05-17

The performance of face recognition systems that use two-dimensional (2D) images is dependent on consistent conditions such as lighting, pose and facial expression. We are developing a multi-view system utilizes three-dimensional (3D) information about the to make more robust these variations. This work describes procedure for constructing database 3D models matching this 2.5D scans which captured from different views, using coordinate invariant properties surface. simplified (x, y, z)...

10.1109/icpr.2004.1334127 article EN Proceedings of the 17th International Conference on Pattern Recognition, 2004. ICPR 2004. 2004-01-01

In this work, we demonstrate a sheathless acoustic fluorescence activated cell sorting (aFACS) system by combining elasto-inertial focusing and highly focused traveling surface wave (FTSAW) to sort cells with high recovery rate, purity, viability. The microfluidic device utilizes particle align in single file for improving accuracy efficiency without sample dilution. Our can effectively focus 1 μm particles which represents the general minimum size majority of applications. Upon...

10.1021/acs.analchem.9b03021 article EN Analytical Chemistry 2019-11-06

Using deep neural networks to solve PDEs has attracted a lot of attentions recently.However, why the learning method works is falling far behind its empirical success.In this paper, we provide rigorous numerical analysis on Ritz (DRM) [48] for second order elliptic equations with Neumann boundary conditions.We establish first nonasymptotic convergence rate in H 1 norm DRM using ReLU 2 activation functions.In addition providing theoretical justification DRM, our study also shed light how set...

10.4208/cicp.oa-2021-0195 article EN Communications in Computational Physics 2022-01-01

Current two-dimensional image based face recognition systems encounter difficulties with large facial appearance variations due to the pose, illumination and expression changes. Utilizing 3D information of human faces is promising handle pose lighting variations. While shape a does not change head (rigid) changes, it invariant non-rigid movement evolution, such as expressions aging effect. We propose surface matching framework take into account both rigid match 2.5D model. The registration...

10.1109/acvmot.2005.40 article EN 2005-01-01

The performance of face recognition systems that use two-dimensional images depends on consistent conditions w.r.t. lighting, pose, and facial appearance. We are developing a system utilizes three-dimensional shape information to make the more robust arbitrary view, For each subject, 3D model is constructed by integrating several 2.5D scans from different viewpoints. A scan composed one range image along with registered 2D color image. engine consists two components, surface matching...

10.1109/acvmot.2005.64 article EN 2005-01-01

10.1016/j.bspc.2008.04.003 article EN Biomedical Signal Processing and Control 2008-06-05

In recent years, physical informed neural networks (PINNs) have been shown to be a powerful tool for solving PDEs empirically.However, numerical analysis of PINNs is still missing.In this paper, we prove the convergence rate second order elliptic equations with Dirichlet boundary condition, by establishing upper bounds on number training samples, depth and width deep achieve desired accuracy.The error decomposed into approximation statistical error, where given in C 2 norm ReLU 3 (deep...

10.4208/cicp.oa-2021-0186 article EN Communications in Computational Physics 2022-01-01

We present a microfluidic device for high-throughput, size-based bacterial sorting from whole blood in non-Newtonian fluids, enabling rapid and simple purification of bacteria more accurate molecular diagnosis bloodstream infection.

10.1039/d1lc00085c article EN Lab on a Chip 2021-01-01

Inertial microfluidics has been proven to be a powerful tool for high-throughput, size-based cell sorting in diverse biomedical applications. In the case of Candida-related sepsis, Candida species and major blood cells (i.e., red white cells) have size distribution 3–5 6–30 μm, respectively. To effectively retrieve majority remove most interfering accurate molecular analysis, inertial micron-sized biological particles with submicron difference is highly desired, but far unexplored till now....

10.1021/acs.analchem.0c03718 article EN Analytical Chemistry 2020-11-16

In this paper, a method to fast classify (Intradural hemorrhage, epidural and cerebral parenchymal hemorrhage) locate the bleeding points by using Singularity Expansion Method (SEM) Backpropagation (BP) neural network optimized genetic algorithm (GA) Sparrow Search Algorithm (SSA) is proposed. simulation model, spot with radius of 3 mm successfully identified approach. The test accuracy in for both bleeding's localization classification are 98.0% 97.4%, respectively. Head phantoms that have...

10.1109/tim.2023.3348908 article EN cc-by IEEE Transactions on Instrumentation and Measurement 2024-01-01

In this paper, Adaboost and SVM are applied to SAR ATR (synthetic aperture radar automatic target recognition) respectively. The performance of these two classifiers is analyzed compared in aspect window with different size. First, PCA (principal component analysis) features selected as feature, then Adaboost.Ml used classify, Experimental results based on MSTAR data sets show that classifier has better robustness than

10.1109/icr.2006.343515 article EN 2006-10-01
Coming Soon ...