Brian D. Bue: [Bio] [CV] [Research Interests] [Publications] [Code] [Teaching]

Brian.D.Bue at rice.edu
Department of Electrical and Computer Engineering
Rice University, 6100 Main St., Houston, TX, 77005
Office: DH 1034
Mobile: 323-638-7684 (gvoice)

Spring 2012 Schedule (CST)

Wed 10:30-12:00PM: ENGI600
Wed 01:00-02:30PM: Group meeting
Wordle generated from my 2009-2011 papers. [wordle generated from my 2009-2011 publications]

Bio

I am a Ph.D. candidate in the Rice University Electrical and Computer Engineering department. My advisor is Professor Erzsébet Merényi of the Department of Statistics.

I was the primary developer (Nov. 2010-Oct. 2011) of the Hyperspectral Image Interpretation and Holistic Analysis Toolkit (Hii-HAT), originally implemented by Dr. David Thompson (PI) and Dr. Lukas Mandrake.

I was a NASA Graduate Student Researchers Program fellow from 2007 to 2010. The title of my GSRP research proposal was "Automatic Labeling of High Dimensional Remotely Sensed Imagery via Semantic Modeling." My NASA technical advisor was Dr. Kiri Wagstaff of the Jet Propulsion Laboratory.

Before coming to Rice, I was a member of the technical staff in the Machine Learning and Instrument Autonomy Group at the NASA Jet Propulsion Laboratory. Prior to JPL, I worked with Dr. Tomasz Stepinski, then of the Lunar and Planetary Institute, now at the University of Cincinnati Department of Geography, on automated landform classification and crater counting by analyzing Mars Orbiter Laser Altimeter (MOLA) data.

I received a M.S. from the Purdue University Computer Science department in 2006, a B.S. in Computer Science and a B.A. in Mathematics from Augsburg College in Minneapolis, MN in 2003.

CV

My CV is available here: [pdf] (last updated: May 04, 2012)

Research Interests

I develop techniques to automatically identify materials represented by high-dimensional "hyperspectral" spectral signatures (pixels). These techniques are largely dependant upon making robust comparisons between unlabeled spectra which we wish to identify, and labeled spectra of known material species. The conditions under which spectra are captured dictate the methodology we use to compare them, as differences in sensor types, environmental conditions, and spatial / temporal effects can significantly alter the manner in which spectra are represented.

When spectra are captured under similar conditions, determining which spectral bands are most relevant (and in what proportions) to the material identification task is crucial. Additionally, spectral bands are often strongly correlated, and suppressing superflous signal content will improve the quality of similarity measurements. The first component of my research develops adaptive, domain-specific spectral similarity measures, learned from a small number of labeled spectral signatures. By replacing traditional task-agnostic similarity measures (e.g., the Euclidean distance) with these learned measures, hyperspectral image segmentation, classification, and spectral library retrieval results can be significantly improved.

When spectra are captured under different environmental conditions or by different sensors, reconcilng these differences is necessary in order to accurately compare them. The second component of my research develops domain adaptation algorithms that automatically reconcile systematic spectral differences between images captured under differring conditions.

Publications

Journal Articles

Conference Proceedings

Code

Teaching

For each Fall semester from 2008 to 2010, I assisted with Prof. Devika Subramanian's "Introduction to Computational Thinking" course.

During Spring semester 2009, I was a teaching assistant for both ELEC502: Artificial Neural Networks with Professor Merényi and COMP540: Machine Learning with Professor Subramanian.

While at Purdue, I was a teaching assistant for Systems Programming Laboratory (C), and Compilers: Principles and Practice, both with Prof. Dennis Brylow (now with the Marquette University Computer Science department).