This website is an electronic resource for news and information about our department.
NEW in ECE
NEWS & INFORMATION:
__________________________________________________________________________________
Dr. Lauren Christopher ~ Receives Awards to begin Research on NanoTechnology System: 3-Dimensional Image Capture with Depth Map Using Fast Focus Nano/Microfluidic Lens and Bayesian Depth from Defocus Image Processing .
Parallel Processing applied to Medical Image Segmentation with IBM Cell Processor
Abstract: The research proposed will begin a new collaboration between experts in nano/microfluidic lens technology and experts in 3-Dimensional Image Processing to produce a feasibility prototype for a 3D image capture system. There are many applications for 3D image capture devices that capture the visible light image plus a depth map. These include robotics, computer vision, ranging data for military applications, vehicle safety systems (collision avoidance), medical endoscopy and potentially consumer electronics. Current 3D range data is acquired by either having multiple imagers (binocular vision) or using radar ranging techniques with frequencies up to coherent light (laser).
We propose using a new single lens method created by two or more images captured by a single imager, but at different focal lengths – called “Depth from Defocus.” This method combines a few multiple natural light images with different focus points to create a depth map, accurate to within about 4% of the distance from the camera. The applications that could benefit from this technique are typically not single stationary images, but movies. In this case, the multiple focal length images must be taken very closely spaced in time. Current glass lens technology requires moving parts, which is far too slow for this application. In addition, the mathematics to generate the depth map is iterative and cannot achieve movie speed with standard hardware. The hypothesis of this research: The combination of a microfluidic lens with high speed algorithm hardware can achieve movie speeds with highly accurate 3D depth information.
Our proposed prototype would characterize the response time of the lens re-focus, and address the speed and accuracy of the algorithm. The traditional glass lens would be replaced by a nano/microfluidic lens, its focus actuated by a piezo-electric subsystem. This will substantially improve the response time of the change in focal length. The iterative depth from defocus algorithm speed can be improved by parallel processing the image(s) data through dedicated hardware. The feasibility demonstration would use programmable logic for the algorithm, and a commercially available microfluidic lens.
The long term objective is to use the RSFG initial research collaboration results to launch a new area of 3D imaging systems research, with funding potential from Darpa for ranging and robotics applications, collaboration with TASI for active vehicle safety applications, and funding from NIH for endoscopic applications.
__________________________________________________________________________________
Parallel Processing applied to Medical Image Segmentation with
IBM Cell Processor
Three dimensional medical images require significant memory and processing power to enable Computer Aided Diagnosis (CAD), planning of radiation therapy, and image-guided surgery. Today these tasks are slow on standard processing platforms, and require significant clinician interaction. Iterative segmentation tasks on large 3D data sets (1-10GByte) are measured in hours, but produce superior and more repeatable results compared to current segmentation techniques. If advanced algorithms were available on parallel processing hardware, throughput of segmentation tasks could be measurably improved for the clinical applications mentioned above.
The IBM Cell Processor is a nine-processor Single Instruction Multiple Data (SIMD) architecture chip, initially developed for 3D image rendering for computer gaming applications. These rendering tasks also exist in 3D medical applications, however the data that is imaged must be segmented properly before rendering. The IBM research team on IUPUI campus has begun research in medical image processing. The team has some initial results on various algorithms for image registration tasks, showing 10x throughput improvement. Image segmentation is on the IBM roadmap for evaluation, but no current projects exist.
This proposal for Purdue Research Faculty Grant is to begin the initial trials of porting existing iterative Bayesian segmentation algorithm to the IBM Cell processor SIMD machine. The study of efficient algorithm parallelization and memory bandwidth optimization will provide initial data which could support ongoing research grants through IBM or independently with NIH.
__________________________________________________________________________________
Diagnosis and Tolerance of Faults and Misbehavior in Dynamic Systems and Networks
Abstract: Faults in large-scale dynamic systems and networks can compromise their functionality in complex ways and, depending on the underlying application, can have devastating consequences or even lead to loss of life. These faults are in general treated as incorrect state or events associated with hardware or software in dynamic systems and networks resulting from functional changes of components, physical interference from the environment and operation error. As the complexity of practical systems continuous to grow, the likelihood of occurrence of faults in these systems increases drastically. Therefore, diagnosis and tolerance of these faults become a crucial task for us in order to build reliable systems, maintain desired commands and operations, and avoid resulting economical
losses.
The purpose of this research is to investigate effective and efficient fault diagnosis and fault tolerance methods in dynamic systems and networks. The long term objectives of this research are to develop theories, build tools, and design algorithms for diagnosis and tolerance of faults that may occur in practical dynamic systems. This would be a significant step in the development of theory, techniques, and software that are able to serve both academia and industry. The research pursued under this proposal has specific aims as follows: (1) develop effective model-based probabilistic fault diagnosis approaches; (2) propose efficient belief propagation algorithms for fault diagnosis in traffic systems; (3) design systematic fault-tolerant approaches for traffic systems and communication
networks.
This research by the support of RSFG will generate novel fault diagnosis and fault tolerance algorithms, approaches, and software packages, with funding potential from NSF, DARPA, and DOT.
__________________________________________________________________________________
PHEV: Plug-in Hybrid Electric Vehicle

L to R~ Kyle Cline, Dr. Sohel Anwar, Senator Dick Lugar, & Dr. Yaobin Chen are shown standing next to the
PHEV: Plug-in Hybrid Research vehicle.
PHEV: Plug-in Hybrid Electric Vehicle
Research is being conducted on the Toyota Prius Hybrid Plug-In vehicle at the IUPUI campus, Richard G. Lugar Center for Renewable Energy . A pair of baby-blue painted 2008 Toyota Prius sedans are being souped up with new equipment and a type of battery that promise to add an additional 60 miles between charges.
__________________________________________________________________________________
The Office of Career Services and Professional Development for the Purdue School of Engineering and Technology is conducting a survey to acquire information on our graduating students. Through this survey we hope to determine what companies are hiring our graduates, what the average salary is for each of the different disciplines within the school, and if internship and cooperative education is making an impact. This information will ultimately be used to improve our services and to provide crucial statistical information for future students.
https://www.et.iupui.edu/careerfair/survey.aspx