Projects


Sponsored by            

Current

Video-based Tracking and Action Analysis
Accurately tracking individual's free movements enables better understanding of a person's activity level, inter-person interaction, habits, etc. In this project, we aim at achieving a precise evaluation of human free movement in 3D space to quantify a person's energy expenditure and emotional status.

  • X. Yuan and L. Kong, Automatic Feature Point Detection for Human Action Tracking in Videos, submitted for review
  • X. Yuan and D. Mohapatra, Removing Complex Shadows from Indoor Videos, submitted for review
Secure Image Processing
Switching to cloud computing is inevitable. While storing and processing medical data on the cloud offer great promise, it is not without challenges. Data security and privacy are major obstacles. Medical data transmitted in the network and processed in the cloud are vulnerable to a broad range of adversaries. Data breaches cost the U.S. health care industry nearly $7 billion annually and ensuring patient confidentiality is a pressing need ever. Secure computing is demanded to expand the horizon of cloud computing. This project focuses on developing novel methods for encrypting the medical images and processing the encrypted data to avoid vulnerability in the transmission and processing of images/videos.

  • M. Gomathisankaran, X. Yuan and *P. Kamongi, Ensure Privacy and Security in the Process of Medical Image Analysis, IEEE Int'l Conf. on Granular Computing, Beijng, China, Dec. 13-15, 2013

SAR Image Analysis and Target Detection
This project focuses on fusing multi-look Synthetic Aperture Radar (SAR) imagery to improve both efficiency and quality of images for detecting rare and small targets. Specifically, we aim at developing novel methods that combine SAR and EO data for better target detection in cluttered background.

  • X. Yuan, I. Ternovskiy, Image reconstruction from sub-apertures of circular spotlight SAR, SPIE Defense + Security, Baltimore, Maryland, USA, May 5-9, 2014
Learning from Large, Imbalanced Data
Data acquisition is usually limited due to policy and economy considerations, and hence the number of training examples of each class varies greatly. It is desirable to understand in what circumstances imbalanced data set affects the learning outcomes and robust methods are needed to maximize the information embedded in the training data set without relying much on user introduced bias. In this project, we study the effects of uneven number of training images for automatic face recognition and proposed a multi-class boosting method that suppresses the face recognition errors by training an ensemble with subsets of examples. By recovering the balance among classes in the subsets, our proposed multiBoost method circumvents the class skewness and demonstrates improved performance.

  • *M. Abouelenien, X. Yuan, Boosting for Learning from Multiclass Data Sets via a Regularized Loss Function, IEEE Int'l Conf. on Granular Computing, Beijing, China, 2013
  • *M. Abouelenien and X. Yuan, Incremental SampleBoost for Efficient Learning from Multi-class Data Sets, SIAM Int'l Conf. on Data Mining, Austin, Texas, USA, May 2-4, 2013
  • *M. Abouelenien and X. Yuan, Study on Parameter Selection using SampleBoost, Flairs, St. Pete Beach, Florida, USA, May 22-24, 2013
  • *M. Abouelenien and X. Yuan, SampleBoost for Capsule Endoscopy Categorization and Abnormality Detection, Int'l Conf. on Adv. Machine Learning Tech. and App., Cairo, Egypt, Dec. 8-10, 2012
  • X. Yuan and *M. Abouelenien, A Boosting Method for Learning from Imbalanced Data toward Improved Face Recognition, Int'l Conf. on Machine Learning and Applications, Boca Raton, Florida, Dec. 12-15, 2012
  • *M. Abouelenien and X. Yuan, SampleBoost: Improving Boosting Performance by Destabilizing Weak Learners Based on Weighted Error Analysis, ICPR, Tsukuba, Japan, November 11-15, 2012
  • *M. Abouelenien and X. Yuan, Improving classification performance for the minotiry class in highly imbalanced dataset using boosting, Int'l Conf. on Computing, Communication, and Networking Technologies, Coimbatore, India, July 26-28, 2012
 

Micromirror Array Simulation and Analysis
Developing a customized micro-mirror array (MMA) is costly and time consuming. Experiments and analysis with physical devices also face combinatorial complexity given the array geometric configurations and light properties. It is therefore desirable to have a computational simulation as a cost-effective and efficient means to conduct exploratory investigations such that appropriate parameters and optical characteristics can be obtained prior to any physical tests. This project aims at developing such a model and simulation platform that allows studies from electrical control to mirror array geometric layout to optical characteristics, and it enables a computational means to evaluate laser-based counter measure performance in future.

  • X. Yuan, *S. Liu, J. Schmidt, I. Anisimov, Micromirror array simulation and far-field diffraction analysis, SPIE Defense + Security, Baltimore, Maryland, USA, May 5-9, 2014
 

Completed

Human Trusted Decision Fusion using Classifier Ensemble and Subgroup Feature Selection (WP AFB, 2011.5 - 2011.8, 2012.5 - 2012.8, 2013.6-2013.8)
Infusing Advanced Sensor Network Research into Cross-disciplinary Undergraduate Education (NSF, 2009.4 - 2011.12)
Computer-aided Diagnosis for Gastrointestinal Bleeding using Wireless Capsule Endoscopy(Texas ARP, 2008.6 - 2010.6)
A New Tool for Economic and Environmental Planning - Expanding the Boundaries of LiDAR (NSF/SGER #0722106, 2007.7 - 2008.6)
US/China Digital Government Collaboration: A New Tool for Economic and Environmental Planning - Expanding the Boundaries of LiDAR (NSF/SGER #0737861, 2007.9 - 2008.8)
Fusing LiDAR and Infrared to Model and Simulate Hydrological Events (ORAU, 2008.7 - 2009.6)