SHORT REVIEW
of recent
publications (beginning from 1995) by Georgii Khachaturov.
You
can download some compressed papers in PS format.
The main question that
provoked the following series of papers was: How a structural
object model should depend on the available routines of low-level
processing?
"Structural Matching to Control the
Low-Level Processing" , (with F. Macias-Rangel), in Proc. of
Congreso International De Computación , CIC2000, 13-17 November 2000,
IPN, Centro de Investigación en Computación, Mexico, pp. 273-284,
2000.
"Knowledge-Based and Processing-Oriented Object
Representation" (with F. Macias-Rangel), in Proceedings
of the Workshop AAPR'2000: Advances in Artificial Perception and
Robotics, Eduardo Bayro Corrochano (Ed.), M-00-26 (CC/CIMAT), Guanajuato,
Mexico, pp. 151-158, 2000.
" Detector-Based Visual Identification of Known Objects ", in Enhanced and
Synthetic Vision 2000, J.G. Verly, Ed., Proceedings of SPIE, Vol. 4023, pp.
269-279, 2000.
Three new issues are presented in them (i) A formalization
of low-level processing (ii) A new object representation (TSG-model)
for Pattern Recognition (iii) A new image processing scheme for
verification if an input image contains an object corresponding to a given TSG-model.
The peculiarity of the processing is that its part representing low-level image
processing works under full control of typical AI search interacting with the TSG-model
of object to be identified.
In other words, structural identification is reduced in the papers to AI search
that manipulates with two kinds of feasible block-operations, the one
represented by detectors (this is a formalization of low-level image
processing) and the one represented by operations with sets (union, intersection,
etc.) of image frame.
The most recent paper of this set extends the technique for the case of
possible occlusions and omission of details during detection.
·
· Papers related to Intelligent
Robot Navigation
The ideas developed
in the following set of papers, perhaps, are of common interest. In
particular, they are related to a challenging application: Planning via Internet
a fastest trip by means of all available timetables of all kinds of transport.
The problem of robot
navigation is represented in those papers as an extension of well known shortest
path problem on graphs. Then, an efficient resolution for this extended
problem is presented. For robotics, this approach leads to a new graph-based
representation of its environment. It gives a new point of view about robot
learning, perception, and other parts of robot architecture and suggests a new
research program for them.
" An Approach to Trip- and
Route-Planning Problems", Cybernetics and
Systems: An International Journal, Vol. 33, pp. 43-67,
2002, Taylor & Francis Group
" An Approach to the Natural Posed Trip
and Route-Planning Problems", in Proceedings of the 2nd
International Symposium on Robotics and Automation, ISRA'2000, Monterrey,
Mexico, pp. 87-92, 2000.
"A Graph-Based Approach To Intelligent Robot
Navigation" in Proc. of Congreso International De Computación , 15-19
November 1999, pp. 453-461, IPN, Centro de Investigación en Computación,
Mexico, 1999.
"Robot
Navigation with an Almost Natural Interface", in Proc. of 10th Irish
Conference on Artificial Intelligence & Cognitive Science (AICS99),
pp. 116-122, University College Cork, Ireland, 1999.
"An
approach to the non-Cartesian navigation", in Proc. of SIARP’99 -
4th Iberoamerican Simp. on Patt. Recogn., pp. 67-78, Havana, Cuba,
1999.
The original goal was a one-step extraction of
"something one-dimensional in picture, disregarding irrelevant
details". It led to a generalization of edgel. "The
3-Frequencies Method" (3FM) was developed for that purpose in
"An Approach to Detection of Line
Elements". In Proc. of the Second Asian Conference on Computer
Vision (ACCV’95), Vol. III, pp.559-563, 5-8 December 1995, Singapore, 1995.
In the following two papers 3FM was applied to a next
step generalization of edgel, and also to extraction of "something
0D".
"Generalization
of Local Elements of Pictures",in Proc. of 30th Intl.
Symposium on Automative Technology & Automation, ISATA-97, Florence,
Italy 16-19 June 1997, pp.257-264, 1997
"Non-matching detection of the local image elements", (with I. Grandi) in
Proc. of TIARP’97 – 2d Iberoamerican Workshop on Patt.
Recogn., pp. 17-27, Havana, 1997,
This set of three papers presents a simple, robust,
and precise method of 3D-pose estimation that can be used, for example, for
spacecraft docking and robotics.
"An Approach to Full Pose Estimation
for an Automatic Control System Based on Vision", (with H.Moncayo)
in Enhanced and Synthetic Vision 1998, J.G. Verly, Ed., Proceedings of SPIE
Vol. 3364, pp.303-313, 1998.
"A Monocular 3D Pose Estimator Based on Fourier Technique", in Proc. of III
Iberoamerican Workshop on Pattern Recongnition, (Memorias del III Taller
Iberoamericano de Reconocimiento de Patrones), March 23-27, 1998, pp.113-129,
ICIMAF (Cuba), CIC IPN, IPN (Mexico), 1998.
"Non-uniformity of a pattern and "the
best" single view 3D pose estimator", Computación y
Sistemas (an international journal in Computer Science, published in CIC IPN,
Mexico), vol II,
no.2 and 3, pp. 114-125, 1999.