(Last Update, August 26, 2003)
Department of Electrical Engineering, University of Kentucky, 453 Anderson Hall, Lexington, KY 40506-0046, Phone: (859) 257-8040, Fax: (859) 257-3092, Email: lgh@engr.uky.edu, URL: http://www.engr.uky.edu/~lgh/
Primary areas of research are pattern recognition and N-D signal
processing. His research program presently has two main branches; automatic
target recognition and 3-D data acquisition. His research includes several
approaches in signal/image detection, discrimination, estimation, registration
and training set selection. Most notable is his work with Synthetic Discriminant Functions (SDFs) for
distortion-invariant optical pattern recognition. He has developed several new
optical pattern recognition schemes. One of these schemes, known as linear
coefficient composite filters is a discrete form of harmonic expansion. He has
extended this research to include multiple distortions based on the discrete
form of multi- parameter harmonic expansions. The resulting filter designs are
numerically efficient for on-line generation. The latest research uses these
rotation-invariant filters in such a way as to reconstruct larger scenes from
smaller, partially overlapping, sub-images. These latest results achieve
numerical efficient operation and are competitive in speed and precision with
manual scene assembly by humans. Other studies of this research include pattern
discrimination, registration, high level morphological operations,
contrast-invariance, fractal colored noise
synthesis/analysis and distortion parameter estimation.
His second area of research is 3-D data acquisition using structured
light techniques. His studies date back to 1980 in this area and include
various algorithms for structuring the light, analysis and reconstruction of
surfaces. In the last year, this branch of his research has received
considerable industrial interest. Research performed but not published yet,
include artificial intelligence (AI) algorithms for active measurement
optimization and sub-pixel accuracy using non-linear filtering of
non-stationary stochastic processes. Most recently he has applied Information
Theory to successive striping methodology to achieve both a deeper
understanding of structured light illumination as well as more efficient
methods. Future research efforts will be the fusion of 3-D data acquisition
with the SDF research to achieve 3-D pattern recognition for AI applications.