Research Areas: 

Graphics Metafile Compositing. Compositing computer graphics metafiles (CGM) is the process of applying various
Boolean operators among potentially overlapped primitive shapes specified within a file designed to create a visual image.
The specific objective of this work is to retrieve multi-layered Windows^® Metafile command records from an image and
translate them into a set of closed contours in a single-layer that delineates the contiguous regions that would be painted
by standard graphics library fill commands. It is largely analogous to a special case of the map overlay problem within
computational geometry and comprises steps that include intersection, search, and boundary traversal algorithms.
However, because many records may overlap in a single region, it is very costly to compute the Boolean operator on
each record sequentially (requiring exponential run-time in the worst case). Here, a fast and robust region scan method
for graphic Boolean operations is presented which processes large numbers of input polygons even within the presence
of erroneous or degenerate data (self overlaps, grazing, etc.).Specifically, an implementation of polygonal object
intersection is presented using a sweep-line algorithm that handles degenerate cases such as parallel or multiple
congruent polygonal object edges. This implementation does not require numerical perturbation to achieve a general
topological position and maintains the time/space complexities of the line segment intersection algorithm. This work has
been successfully implemented in the context of an embroidery design automation application.Furthermore, since the
algorithms reduce the size of a Windows metafile significantly and without loss of image quality, they promise to be
useful in augmenting methods for compressing vector file images and may have other useful engineering applications. 

Algorithms for Vector Graphic Optimization and Compression. Many existing graphic compression methods such as
GIF, PNG, and JPEG etc. do not work well for the compression of vector graphic files because these compression methods
are based on compressing bitmap graphics with either statistical modeling algorithms or encoding algorithms.After
compression, the file format is changed and the resolution-independent nature is also lost.A CGM compositing algorithm
is proposed here for vector graphic file compression and optimization.In this algorithm there is no loss in the quality of
the resulting image.Furthermore, the file format is not changed, which means that it is resolution independent.This
method has two major advantages: First, applications do not have to call other software that supports a new file format as
they do with most graphics compression applications.Second, after compositing, the file is saved back to its original
vector file format, thus maintaining its vector features so that geometric transformations such as rotation and scaling may
be applied subsequently.Other compression software changes the format of the image file. The composting algorithm
sequentially takes polygons with filling mode and color attributes as input and then outputs a set of consistently formed
non-overlapping polygons.The input polygons need not necessarily be regular polygons, i.e. polygon vertices may be
specified in any order (clockwise or counter-clockwise) and the polygon itself might be self-overlapped.The output is
order-specified, i.e. the outline polygons are in counter clockwise order and any contours indicating holes are specified
in a clockwise order.After compression with our CGM compositing algorithm, redundant records are removed and the
output records are recreated in a consistent order.Therefore the vector graphic is optimized as a byproduct. 

Structural Indexing in Automated Embroidery. This area of work involves the implementation of new structural indices
(incorporating results of our work on multidimensional indexing using B* and G tress and fast Voronoi diagram
computation) that allow effective categorization and identification of varied 2-dimensional vector features.  Specifically,
methods that index and allow efficient search and retrieval of regular shapes as well as singular of regions are being
developed. Additional indices are being developed to allow various levels of comprehensive shape recognition.  All
of the resulting indices and methods will then be combined into an expert system that will allow interpretation of artwork
for automatic stitch placement mechanisms and reproduction using commercial embroidery equipment. 

New Stitch Generation Engine.This area of work comprises the development of a new stitch generation software engine
that will more accurately and robustly compute sequences of stitch locations based upon primitive vector input
commands.  Specific care in handling degenerate cases is very important in this work (and is a shortcoming in existing
stitch generation engines currently in use).  Previous work has yielded good results for fundamental stitching primitives
(e.g., fills and columns). The focus now is on enhancing that implementation and adding support for more advanced
primitives (e.g., trapunto fills, advanced pattern fills, etc.).  In addition, new vector primitive commands such as Bezier
curves will be implemented providing another benefit over existing stitch generation engines. 

General Student Models for Intelligent Tutoring Systems. As technology-based learning environments such as
Intelligent Tutoring Systems and Adaptive Educational Hypermedia become more prevalent, students may interact with
many such learning environments during their studies.  Throughout its interaction with a user, a learning environment will
gain a great deal of knowledge about the student (through its student model).  However, after the user completes the
training, all of the information about the student will be lost, much of which could be useful for other learning
environments the student will encounter in the future.  So this research attempts to find out if there is a way to effectively
store information from a learning environment’s student model so that other learning environments the student encounters
in the future can utilize this information to further adapt its instruction to the needs of the student.  One answer is to use a
General Student Model (GSM): a centralized data base that stores individual learning environment data and student
information which may be retrieved and then used to enhance the learning environment’s own student model, which in
turn further adapts the learning environment to the needs of the student.  A GSM has been developed and is being tested
by having learners use three separate systems which incorporate the GSM to share student information: a Student Attribute
Test, an Educational Adaptive Hypermedia System, and an Intelligent Tutoring System.  Users are being split into two
groups: a GSM-enhanced (treatment) group in which the learning systems will be connected to the GSM and a
Non-GSM-enhanced (control) group which is not connected to the GSM.  An Adaptive Hypermedia Engine has been
created which delivers tailored web pages to the user based on learner attributes stored in the GSM, as well as the student’s
interaction with system.  An Intelligent Tutoring System has been constructed to help teach elementary German for travelers.
It collects field dependence and impulsivity data from the GSM and collects knowledge of German data from the subject.
The following questions will be answered after the system has been run and the results analyzed: Do students who learn
from GSM-enhanced systems have better mastery of the subject?  Do they learn the same material more quickly?  Do they
spend more time exploring the learning environments?  Do they explore more of the learning environment?  Are they more
satisfied with their learning experience?  Do they have more confident in their abilities?  Are they more likely to continue
studying the subject matter?  A test system is up and running. 

Educational Robotics. We are working on the design of low-cost autonomous robots that use Bluetooth, Zigbee, and RFID
wireless technologies for communication. 

Recent Research Sponsors: NSF; SPIR; Soft Sight, Inc.

Recent Publications:

Abraham L. Howell, Roy T.R. McGrann, Richard R. Eckert, "Using ZigBee to Control a Swarm of Low Cost RFID Foraging
Robots," The 4th International Conference on Cybernetics and Information Technologies, Systems and Applications:
(CITSA 2007), jointly with The 5th International on Computing, Communications and Control Technologies: CCCT 2007),
Orlando, Florida, USA (July, 2007).

Mingkui song, Richard R. Eckert, David A. Goldman, "Graphics Metafile Compositing: A Robust Region Scan Algorithm
for Sequenced Planar Data," The UK Chapter of the Eurographics (EGUK) Conference on the Theory and Practice of
Computer Graphics2007 (TPCG07), University of Wales, Bangor, UK (June, 2007).

Howell, A., McGrann, R., Way, E., Eckert, R., and Sayama, H.,  "Using RFID and a Low-Cost Robot to Evolve Foraging
Behavior," 2006 GECCO Conference, Seattle, WA, (July 8-12, 2006).

Mingkui Song, Richard R. Eckert, David A. Goldman, “Algorithms for Vector Graphic Optimization and Compression”, CGI
2006, LNCS 4035, Hangshou, P.R. China, pp. 665 – 672, 2006 (June, 2006).

David Goldman, Mingkui Song, Richard R. Eckert, "Metafile Compositing for Automated Embroidery Design Generation,"
International Conference on Graphics, Vision and Image Processing (GVIP-05), Cairo, Egypt (December 2005).