Computer
Vision Group at Florida Tech
Research Students
Roman Filipovych
Wei Liu
Anand Mehta
Jaron Blackburn
Bridgette Wiley
Past Graduate Students
Arturo Donate (DOE GANN Ph.D. Fellow, Florida State University)
Gary Dahme
Research
Projects
Analyzing
Non-Rigid Textured Surfaces
Our
main goal in this study is to investigate new algorithms
for
modeling and classifying images of non-rigid deforming texture
surfaces. Traditional methods for texture modeling are usually based on
local texture measurements performed on fronto-parallel planar
surfaces. However, in the past few years, the problem of analyzing
non-rigid textures has received growing attention from the computer
vision community. Examples include works on dynamic textures, non-rigid
structure from motion, and non-rigid texture classification. In our
group, we are investigating new algorithms for classifying video
sequences of patterned surfaces undergoing significant levels
of
curvature-induced distortion. This is a challenging problem
as
the appearance of local texture can vary significantly due to geometric
warping caused by surface curvature.
Papers:
Roman
Filipovych and Eraldo
Ribeiro.
Learning Repetitive Patterns for Classifying Non-Rigidly
Deforming Texture Surfaces. International
Conference on Image Analysis and Processing (ICIAP),
Modena, Italy, Sept. 2007. (pdf)
(Bibtex
citation).
Roman
Filipovych and Eraldo
Ribeiro.
Learning Basic Patterns from Repetitive Texture Surfaces
Under Non-Rigid Deformations. International
Conference on Image Analysis and recognition (ICIAR),
Montreal, Canada, August 2007. (pdf)
(Bibtex
citation).
Image-Based
Biofouling Characterization
In
this project, our goal is to develop recognition algorithms for
for the automation of traditional antifouling coating
evaluation
procedures in field testing. The automation of these testing sites is
of primary importance to the current combinatorial antifouling coating
research developed by the U.S. Navy. However, at the field testing
stage, the analysis of the effectiveness of these coatings is mostly
accomplished by human visual inspections. Image-based
inspection
software will significantly improve the accuracy and speed of
the
testing procedures of antifouling coatings. The current focus of our
projects is on the characterization of fouling organisms such as
barnacles, tubeworms, encrusting and arborescent bryozoans, mollusks,
and sponges.
Papers:
Franck
Casse, Eraldo
Ribeiro,
Abdullah Ekin, Dean C. Webester, James A. Callow and Maureen E. Callow.
Laboratory Screening of Coating Libraries for Algal Adhesion.
Biofouling,
pp.1-10, May, 2007.
Automatic
Pollen Recognition
Pollen
analysis (palynology) is employed in a wide variety of industrial, 
health-related,
and forensic applications. This project seeks to conduct research on
object recognition for pollen identification and can apply equally well
to modern and fossil grains. Pollen
analysis also plays an important role in climate change modeling.
Fossil pollen spectra recovered from lake and ocean sediments, or ice
cores, are an important source of proxy data that allow modeling of
past climates and regional net primary productivity. This work is
performed in collaboration with the Neotropical
Paleoecology Research Group
at Florida Institute of Technology.
health-related,
and forensic applications. This project seeks to conduct research on
object recognition for pollen identification and can apply equally well
to modern and fossil grains. Pollen
analysis also plays an important role in climate change modeling.
Fossil pollen spectra recovered from lake and ocean sediments, or ice
cores, are an important source of proxy data that allow modeling of
past climates and regional net primary productivity. This work is
performed in collaboration with the Neotropical
Paleoecology Research Group
at Florida Institute of Technology.Papers:
Gary
Dahme, Eraldo
Ribeiro,
and Mark Bush. Spatial Statistics of Textons. International
Conference of Computer Vision Theory and Applications -
VISAPP, Setubal, Portugal, 2006. (pdf)
(Bibtex
citation)
Sponsored
by:
Office
of Naval Research
Combining
Visual Cues for Object Recognition
Recent
solutions to object classification are based on single visual cue
measurements. Psychophysical evidence suggests that humans use multiple
visual cues to accomplish recognition. In this work, we address the
problem of integrating multiple visual information for object
recognition. We propose a new probabilistic integration model of
multiple visual cues at different spatial locations across the image.
Our cue integration framework can be used to classify images of
objects.
Papers:
Roman
Filipovych and Eraldo
Ribeiro.
Probabilistic Combination of Visual Cues for Object Classification. 3rd
International
Symposium on Visual Computing.
Lake Tahoe, Nevada, USA, November 2007 (pdf) (Bibtex
citation).
Human
Motion Recognition and Tracking
In
this work, we are developing novel methods for the recognition and
understanding of human motion from videos. Our main focus is on
Bayesian probabilistic models that allows us to learn motion
representations from a set of training videos. In our initial model, a
video sequence is represented with a sparse set of spatial and spatial
temporal
features by extracting static and dynamic interest points.
This model encodes the spatial-temporal relationships between
the
dynamics of the motion and the appearance of individual poses. Future
directions of investigation include the study of interactions between
humans and objects, gestures, and the interaction between groups of
people.
features by extracting static and dynamic interest points.
This model encodes the spatial-temporal relationships between
the
dynamics of the motion and the appearance of individual poses. Future
directions of investigation include the study of interactions between
humans and objects, gestures, and the interaction between groups of
people. Papers:
Roman
Filipovych and Eraldo
Ribeiro.
Combining Models of Pose and Dynamics for Human Motion Recognition. 3rd
International
Symposium on Visual Computing.
Lake Tahoe, Nevada, USA, November 2007 (pdf)(Bibtex
citation).