Dr. Yingzi (Eliza) Du
Director, Biometrics and Pattern Recognition Laboratory
Assistant Professor, Department of Electrical and Computer Engineering
Home Teaching Research Publications News
Research Projects
(2006~Present)
1. Non-cooperative Iris Recognition System
Abstract: Currently, there is no iris recognition system that can perform positive human identification in video surveillance. The challenge is to identify a non-cooperative subject. The quality of an iris image obtained from a non-cooperative subject is often poor. For a non-cooperative subject who may be facing away from the camera, off-angle iris images will often be captured with motion blur or out-of-focus. In this project, we propose an efficient iris recognition system for non-cooperative user identification. It will be based on iris video images instead of an individual image frame. It will use selected features with sufficient quality to generate a multi-resolution template for multi-stage matching. This system can also be adopted for cooperative user identification. Since this system can work with low quality images, it can help to stretch the range of regular iris recognition systems.
Sponsors: ONR, DOD, and IUPUI Research Support Funds Grant (RSGF).
2. Transforming traditional iris recognition systems to work on non-idea situations
Traditional Iris Recognition System
Our Proposed Iris Recognition System
Abstract: Under a non-ideal situation, the image quality may vary. As a result, the traditional iris recognition systems would not work well. However, these kinds of iris recognition systems have been widely deployed in law enforcement and homeland security. It will be desirable to transform the traditional systems to perform in non-ideal situations without a costly update. In this project, we propose a method that upgrades the traditional iris recognition system to work on non-ideal situations. The proposed method takes into consideration not only the effect of image quality, but also the segmentation accuracy. It employs video-based image processing techniques to quickly identify and eliminate the bad quality images from iris videos for further processing. The proposed method is tested on public databases using in-house recognition algorithms and also evaluated using a commercialized system. The research results show that the proposed methods can be used to improve the performance of iris recognition systems in a non-ideal situation.
Collaborators: Dr. James Matey, US Naval Academy
Sponsors: NIJ, ONR, DOD, and IUPUI Research Support Funds Grant (RSGF).
3. Workload Evaluation
Prediction of a person’s level of fatigue, and therefore measurement of their fatigue response is an important tool in predicting and reducing user error. Measurement of eye movement, dilation, and blink rate has been previously correlated with fatigue, mental workload, and other cognitive processes in the literature.
Collaborators: Dr. Jin-Hua She, Tokyo University of Technology
Sponsors: IUPUI International Development Funds (IDF) and Japan Society for the Promotion of Science.
4. Video based Traffic Monitoring and Classification System
The proposed system
Abstract: Video image processing based vehicle tracking and traffic monitoring provides several advantages over traditional approaches. One of the challenges lies in the robust segmentation and region tracking. The traditional approaches have assumed 1) the region of the object of interest is uniform and homogeneous; 2) adjacent regions should differ significantly. These are often not true for a vehicle object. In this paper, we propose a dynamic content based image segmentation method for vehicle tracking and traffic monitoring. Only initial lane information is needed for camera calibration. The system will automatically detect the direction of the traffic flow for vehicle detection and traffic monitoring.
Collaborators: Dr. Shuo Li (INDOT), Dr. Yi Jiang (Purdue Univ.), Dr. Tommy Nantung (INDOT), and Dr. Samy Noureldin (INDOT)
Sponsor: INDOT
5. Development of Hardware Image Processing Cores
Abstract: The current traffic monitoring system in Project 4 is implemented in MATLAB. To improve the speed and efficiency, it would be desirable to use FPGA implementation for image processing, analysis and computer vision systems. The research challenge is not only in the implementation of image processing algorithms in hardware but also characterizing the performance of FPGAs for image-processing applications. In this project, we will develop a comprehensive library that will contain information such as clocking rates of image processing IP cores, memory bandwidth requirements, number of logic cells used by the IP cores, on-chip and offchip memory requirements and latency and throughput of the hardware image processing IP cores. This library will be the basis for the parameters of the ImagePredicate operator in the SYMBIOTE system and will be used by the query planner, static query optimizer and dynamic query optimizer to determine how much and what portion of the queries should be migrated to hardware rather than software depending on the unused resources on the SYMBIOTE node. Moreover, we will explore existing image processing IP cores such as and investigate ways to integrate such IP cores in the SYMBIOTE system.
Colaberators: Dr. John Lee and Dr. Yuni Xia (IUPUI)
Sponsor: NSF
6. Computer-Assisted Identification of Celiac Ganglia During Endoscopic Ultrasound-Guided Celiac Plexus Neurolysis (EUS-CPN)
The proposed CAD Interface
Abstract: Patients suffering from pancreatic cancer or chronic pancreatitis experience debilitating pain severely impairing the patients’ ability to perform normal daily activities. To help patients become relatively pain-free and thus regaining their mobility, doctors may perform a celiac plexus neurolysis (CPN) procedure which involves blocking the nerves of the celiac ganglia. Endoscopic ultrasound-guided celiac plexus neurolysis (EUS-CPN) is an innovative and minimally invasive approach to performing a CPN using an endoscope. The technique of EUS-CPN was first reported in 1996 and is available in university hospitals; however, inability to reliably identify the celiac ganglia, or nerve bundles, has limited its widespread use in the United States. In this research, we propose using a computer aided method to assist the physician in automatically detecting and identifying the celiac ganglia thus increasing the success of precise injections of the nerves during the EUS-CPN. The computer aided method will provide the physician operator with a user-friendly interface allowing assistance in determining the likelihood that an area of interest is a celiac ganglia bundle versus other internal features such as a lymph node.
Collaborator: Dr. Julia LeBlanc (School of Medicine, Indiana Univ.)
Sponsor: IUPUI
7. Hyperspectral Imaging
(a) (b) (c)
Example of Hyperspectral image and signatures of the objects. (a) Cuprite AVIRIS image scene. (b) Spatial positions of five pure pixels of five pure pixels corresponding to minerals: alunite (A), buddingtonite (B), calcite (C), kaolinite (K), and muscovite (M). (c) Five mineral reflectance spectra and background signature.
Abstract: One of great challenges in unsupervised hyperspectral target analysis is how to obtain unknown target knowledge directly from the data. This project is to automatically detect the targets of interest using unsupervised target analysis.
Collaborator: Dr. Chien-I Chang (Univ. of Maryland Baltimore County)
Sponsor: IUPUI
8. Ca2+ Spark Detection and Classification
Proposed model of Ca2+ sparks in a cell image
The model for cell image estimation
The model to detect the Ca2+ spark events
Diagram of the proposed Ca2+ sparks event detection
Abstract: It is evidenced that Ca2+ plays a critical role in cell activities. Understanding and analysis of a Ca2+ sparks event can provide critical information in cell communications. It is known that a Ca2+ spark spatial spread in a local area can be modeled by a Gaussian distribution. There are some available methods in Ca2+ spark detection, but little research has been done to quantitatively analyze and classify the Ca2+ sparks into single or multiple release events. In this paper, a new method based on Gaussian-Mexican Hat Wavelet is proposed to automatically classify Ca2+ sparks into single release events or multiple releases events. Then Ca2+ sparks from single release events are further categorized into different energy levels.
Collaborator: Dr. Martin F. Schnieder (Univ. of Maryland School of Medicine)
Future Partnership and Sponsorship
We are always open for new partnerships. We would like to work and collaborate with academia, government, industry, and other organizations. Please feel free to contact Dr. Du if you are interested.
------------------------------------------------------------
Contact:
Dr. Yingzi (Eliza) Du
Assistant Professor
Office: SL 164 E
Phone: (317) 278-2276
Fax: (317) 274-4493
Address:
Department of Electrical and Computer Engineering
Purdue School of Engineering and Technology
Indiana University-Purdue University Indianapolis
723 W. Michigan Street, SL 160
Indianapolis, IN 46202
Sectary Phone: (317) 274-9726
