Pixel Statistics and False Alarm Area in Genetic
Programming for Object Detection
M. Zhang, P. Andreae, and M. Pritchard. Applications of Evolutionary Computing,
EvoWorkshops2003: EvoBIO, EvoCOP, EvoIASP,
EvoMUSART, EvoROB, EvoSTIM, volume 2611 of LNCS, page 455--466. University of Essex, UK, Springer-Verlag, (14-16 April 2003)
Abstract
A domain independent approach to the use of genetic
programming for object detection problems. Rather than
using raw pixels or high level domain specific
features, this approach uses domain independent
statistical features as terminals in genetic
programming. Besides position invariant statistics such
as mean and standard deviation, this approach also uses
position dependent pixel statistics such as moments and
local region statistics as terminals. Based on an
existing fitness function which uses linear combination
of detection rate and false alarm rate, we introduce a
new measure called "false alarm area" to the
fitness function. In addition to the standard
arithmetic operators, this approach also uses a
conditional operator 'if' in the function set. This
approach is tested on two object detection problems.
The experiments suggest that position dependent pixel
statistics computed from local (central) regions and
nonlinear condition functions are effective to object
detection problems. Fitness functions with false alarm
area can reflect the smoothness of evolved genetic
programs. This approach works well for detecting small
regular multiple class objects on a relatively
uncluttered background.
%0 Conference Paper
%1 Zhang:evowks03
%A Zhang, Mengjie
%A Andreae, Peter
%A Pritchard, Mark
%B Applications of Evolutionary Computing,
EvoWorkshops2003: EvoBIO, EvoCOP, EvoIASP,
EvoMUSART, EvoROB, EvoSTIM
%C University of Essex, UK
%D 2003
%E Raidl, Günther R.
%E Cagnoni, Stefano
%E Cardalda, Juan Jesús Romero
%E Corne, David W.
%E Gottlieb, Jens
%E Guillot, Agnès
%E Hart, Emma
%E Johnson, Colin G.
%E Marchiori, Elena
%E Meyer, Jean-Arcady
%E Middendorf, Martin
%I Springer-Verlag
%K algorithms, applications, computation, evolutionary genetic object programming, recognition
%P 455--466
%T Pixel Statistics and False Alarm Area in Genetic
Programming for Object Detection
%V 2611
%X A domain independent approach to the use of genetic
programming for object detection problems. Rather than
using raw pixels or high level domain specific
features, this approach uses domain independent
statistical features as terminals in genetic
programming. Besides position invariant statistics such
as mean and standard deviation, this approach also uses
position dependent pixel statistics such as moments and
local region statistics as terminals. Based on an
existing fitness function which uses linear combination
of detection rate and false alarm rate, we introduce a
new measure called "false alarm area" to the
fitness function. In addition to the standard
arithmetic operators, this approach also uses a
conditional operator 'if' in the function set. This
approach is tested on two object detection problems.
The experiments suggest that position dependent pixel
statistics computed from local (central) regions and
nonlinear condition functions are effective to object
detection problems. Fitness functions with false alarm
area can reflect the smoothness of evolved genetic
programs. This approach works well for detecting small
regular multiple class objects on a relatively
uncluttered background.
@inproceedings{Zhang:evowks03,
abstract = {A domain independent approach to the use of genetic
programming for object detection problems. Rather than
using raw pixels or high level domain specific
features, this approach uses domain independent
statistical features as terminals in genetic
programming. Besides position invariant statistics such
as mean and standard deviation, this approach also uses
position dependent pixel statistics such as moments and
local region statistics as terminals. Based on an
existing fitness function which uses linear combination
of detection rate and false alarm rate, we introduce a
new measure called {"}false alarm area{"} to the
fitness function. In addition to the standard
arithmetic operators, this approach also uses a
conditional operator 'if' in the function set. This
approach is tested on two object detection problems.
The experiments suggest that position dependent pixel
statistics computed from local (central) regions and
nonlinear condition functions are effective to object
detection problems. Fitness functions with false alarm
area can reflect the smoothness of evolved genetic
programs. This approach works well for detecting small
regular multiple class objects on a relatively
uncluttered background.},
added-at = {2008-06-19T17:46:40.000+0200},
address = {University of Essex, UK},
author = {Zhang, Mengjie and Andreae, Peter and Pritchard, Mark},
biburl = {https://www.bibsonomy.org/bibtex/222325fde01cccf94d8e35545b2e7ac7d/brazovayeye},
booktitle = {Applications of Evolutionary Computing,
EvoWorkshops2003: Evo{BIO}, Evo{COP}, Evo{IASP},
Evo{MUSART}, Evo{ROB}, Evo{STIM}},
editor = {Raidl, G{\"u}nther R. and Cagnoni, Stefano and Cardalda, Juan Jes\'us Romero and Corne, David W. and Gottlieb, Jens and Guillot, Agn\`es and Hart, Emma and Johnson, Colin G. and Marchiori, Elena and Meyer, Jean-Arcady and Middendorf, Martin},
email = {mengjie@mcs.vuw.ac.nz},
interhash = {ffb60932ebe20b3fcd4ba775e6ef5ca3},
intrahash = {22325fde01cccf94d8e35545b2e7ac7d},
keywords = {algorithms, applications, computation, evolutionary genetic object programming, recognition},
month = {14-16 April},
notes = {EvoWorkshops2003},
organisation = {EvoNet},
pages = {455--466},
publisher = {Springer-Verlag},
publisher_address = {Berlin},
series = {LNCS},
timestamp = {2008-06-19T17:55:32.000+0200},
title = {Pixel Statistics and False Alarm Area in Genetic
Programming for Object Detection},
volume = 2611,
year = 2003
}