|
|
THE RECOGNITION TOOLKIT
is a commercial software package produced by RSI to assist anyone
who wants to develop object recognition applications. This
package is an add-on module to image processing tools on both PC
Windows and
UNIX environments. The Recognition Toolkit provides
feature extraction, feature evaluation, learning capabilities, and
classifiers to ease the construction of tailored image object
recognition systems. The software was developed in C++ to
maximize flexibility and take advantage of object oriented design
benefits. |
|
The UNIX version of the
Recognition Toolkit has been integrated as an add-on module to the Amerinex Applied Imaging (AAI) KBVision
System. The PC version for Windows operates
with the AAI/ADCIS
Aphelion image
processing system. The Recognition Toolkit contains routines for
advanced feature extraction, feature evaluations, learning and
training, and classification. These routines (called Tasks in the
UNIX version and Operators in the Windows version) allow a
developer to easily evaluate the features and attributes which are
extracted from imagery and create tailored classifiers. |
|
Demonstration graphs
(UNIX version) and Basic Macros (Windows version) assist the user
in combining the modules for rapidly prototyping automated
recognition solutions. Some useful macros (RSI
Utilities) are available for download for Aphelion users. |
Almost all of the Tasks from the
UNIX version are available as Operators in the Windows version of the
Recognition Toolkit. Some Tasks were eliminated because the
functions would be accomplished in a different manner on the PC system
due to differences in the image processing support environments.
A few new Operators were added to the Windows version to retain all
the capabilities in the UNIX system and extend some of them. In
particular Operators were added to: create tiled regions, develop
statistics on tiled regions, merge lines and merge boxes. All of
these basic evaluation, learning and training, and classification
Operators are fundamentally the same with some small parameter
variations.
The Recognition
Toolkit provides a capability for pattern recognition on continuous
measurement data, discrete data, or symbolic information. It contains
a series of Tasks
for classification, training, feature evaluation, clustering, coding,
and extraction. Twenty Tasks form the pattern recognition core of the
Recognition Toolkit for the KBVision System, and five additional
advanced extraction and demonstration Tasks provide assistance in
complex recognition development. The core Task set includes four
classifiers, four training Tasks, an automatic natural clustering
routine, five feature evaluation and analysis Tasks, and six utility
Tasks for preprocessing and building multi-stage alternative
classifiers.
|
|