Amar Raheja

Research Interests

My broad research interests are in the following areas:

• Biomedical Imaging and Image Processing
• Computer Vision
• Applications of soft computing methods such as neural networks, fuzzy logic, genetic algorithms etc.
• Graphics and Visualization

External Grant and Contracts

Current Support

Agricultural Research Initiative
Characterization of seeds post germination software using image analysis of digital images captured by an automated robotic camera on rails (ARI 11-4-189-11)
Role: PI ; $70,487 ; March 2011 - March 2012

National Science Foundation
Robust Nonlinear Adaptive Control of Unmanned Aerial Vehicles Using Neural Networks (NSF REU 1102382)
Role:Co-PI (PI: Dr. Subodh Bhandari); $360,000 ; August 2011 - August 2014

Past Support

Agricultural Research Initiative
Automated germination recognition and characterization software for seed quality assessment (ARI 10-4-181-11)
Role: PI ; $39,885 ; March 2010 - March 2011

National Science Foundation
IDBR: Excitation-Emission Matrix Fluorescence Detection for Capillary Electrophoresis (NSF DBI 0754837)
Role: Co-PI (PI: Dr. Timothy Corcoran); $177,564 ; May 2008 - June 2010

Agricultural Research Initiative
Development of recognition software as an initial step to automate the assessment of seed quality (ARI 09-04-172)
Role: PI ; $127,838 ; January 2009 - January 2010

L-3 Communications/Interstate Electronics Corporation
Advanced 3-D Locator Base Station Software (Subcontract of Advanced 3-D Locator for First Responders)
Role: PI ; $78,400 ; January 2007 - June 2008

National Science Foundation
Acquisition of a workstation network for research in parallel and distributed computing (NSF MRI 0321333)
Role: Co-PI (PI: Dr. Hairong Kunag); $159,658 ; September 2003 - August 2005

National Textile Center (US Department of Commerce)
Fuzzy Forecasting Model for Apparel Sales (NTC S01-PH10)
Role: PI; $300,000 ; June 2001 - May 2004

National Textile Center (US Department of Commerce)
Haptic Simulation of Fabric Hand (NTC S00-PH08)
Role: Co-PI (PI: Dr. Muthu Govindaraj); $400,000 ; June 2000 - May 2003

Collaborators (Current and Past) and Resarch Projects

Dr. David Still   Cal Poly, Pomona

Planting high quality seed is the basis by which agriculture remains profitable. Every seed lot sold commercially has had an assessment of seed quality which required germinating seeds under multiple environmental conditions. Final germination under benign (control) conditions indicates the potential of the seed lot, but a sensitive indicator to seed germination is provided by germination rate, particularly if evaluated under various stressful environments which reduce germination capacity. Typically seed companies and commercial seed testing laboratories evaluate thousands to tens of thousands of seed lots each year. Further, in order to discover the genetic determinants of seed germination under environmental stress and the release from dormancy, researchers typically work with large numbers of genetic populations, each of which must be evaluated under a multitude of environmental conditions. A bottleneck exists in evaluating seed germination both for commercial and research purposes. The research we are collaborating on seeks to develop software by which the evaluation of seed germination may be automated by developing algorithms that will count seeds in a Petri dish using an image of the seeds and to develop algorithms that differentiate a germinated seed from an not-germinated seed from images of these seeds taken at various time intervals.

Dr. Timothy Corcoran   Cal Poly, Pomona

Fluorescent labels data collected into a data cube with dimensions excitation × emission × time. Using the method of parallel factor analysis, the abundance of each smoothly-varying spectra of typical fluorescent labels can be extracted as a time series, even in the presence of unknown interferents. Compression of the data cube to as little as 1% of its original size via well chosen multidimensional hybrid wavelet transforms, data analysis accelerates by factors as large as 50 without compromising accuracy. The data acquisition software will be developed using LabView, a powerful programming environment well-suited to instrument data acquisition and the parallel factor analysis implemented in some high level programming language.

Dr. Wely Floriano   Cal Poly, Pomona

Cassandra: a new tool for virtual ligand screening. Cassandra will be a structure-based molecular design computational tool that researchers can easily use to identify/design potential therapeutic compounds for any target protein with known or predicted 3D structure. This tool will allow researches to seamlessly step through different structure-based molecular design applications, from binding site recognition and affinity profiling, to pharmacophore-based searches and genome data mining. This will shorten the training period for new researchers, make our in-house tools available for other laboratories and research groups to use and validate, and will also provide numerous new opportunities for research and educational use.
Currently, QT is being used to develop an intutive multi-platform GUI for Cassandra.

Dr. Dennis Livesay   Univ. of North Carolina, Charlotte

Catalytic site prediction from sequence and structure: This research is investigating ways to predict specific catalytic residues from sequence and structure. We are using machine learning and soft computing techniques to predict catalytic sites solely from sequence-derived information (i.e. alignment conservation, phylogenetic motifs and predicted secondary structure). At the same time, we are investigating structure-based predictions schemes as well. For example, we are currently studying catalytic residue predictions from network models, which recast protein structures as graphs.

Dr. Koushik Adhikari,   Kansas State University, Manhattan

Create a software tool based on neual networks and genetic algorithms to predict the sensory analysis of different food products such as meat, cheese and ice cream. Sensory data sets are complex because they involves human senses and judgments, which can be considered unstructured as opposed to instrumental data. One of the goals of this research is to provide improved understanding of the relationship between specific characteristics of variables that make a desirable and marketable product. This software will also provide the stepping stool to the ultimate goal of reverse engineering of consumer products coming up with the values of independent variables, given the desired values of the dependent variables.