Conference PaperPDF Available

Investigating the effects of ensemble classification on remotely sensed data for land cover mapping

Authors:
Bolanle Abe at Tshwane University of Technology
Bolanle Abe
Anthony Gidudu at Makerere University
Anthony Gidudu
Tshilidzi Marwala at University of Johannesburg
Tshilidzi Marwala
Download full-text PDF
Read full-text
Download citation

Abstract

Ensemble classification involves consulting experts in taking final decision in classification process. The idea is to improve classification accuracy when compared to their single classifier counterpart. The system is used in remote sensing imagery to obtain information about Land cover. Major challenges associated with classification accuracy include design procedure of classifier, choice of training sets from dataset and information conveyed to the algorithm. Superiority of different classification approaches employed depends on selected dataset and the strategy used during designing phase of each classifier. However, in ensemble approach, there is no definite number of classifiers that should take part in decision making. This study exploits feature selection technique to create diversity in ensemble classification. Results obtained show that for ensemble approach, there is no significant benefit in having many base classifiers. The outcome should reveal how to design best ensemble using feature selection approach for land cover mapping.
Content uploaded by Anthony Gidudu
Author content
All content in this area was uploaded by Anthony Gidudu on Jul 17, 2021
Content may be subject to copyright.
,19(67,*$7,1*7+(())(&762)(16(0%/(&/$66,),&$7,21215(027(/<
6(16(''$7$)25/$1'&29(50$33,1*
%RODQOH$EH
D
$QWKRQ\*LGXGX
D
DQG7VKLOLG]L0DUZDOD
E
C
5EJQQNQH'NGEVTKECN+PHQTOCVKQP'PIKPGGTKPI7PKXGTUKV[QHVJG9KVYCVGTUTCPF,QJCPPGUDWTI5QWVJ#HTKEC
D
(CEWNV[QH'PIKPGGTKPIVJG$WKNV'PXKTQPOGPV7PKXGTUKV[QH,QJCPPGUDWTI5QWVJ#HTKEC
#DGAVQNWNQRG"[CJQQEQO#PVJQP[)KFWFW"IOCKNEQOVOCTYCNC"IOCKNEQO
$%675$&7
(QVHPEOHFODVVLILFDWLRQLQYROYHVFRQVXOWLQJH[SHUWVLQWDNLQJ
ILQDO GHFLVLRQ LQ FODVVLILFDWLRQ SURFHVV 7KH LGHD LV WR
LPSURYH FODVVLILFDWLRQ DFFXUDF\ ZKHQ FRPSDUHG WR WKHLU
VLQJOH FODVVLILHU FRXQWHUSDUW 7KH V\VWHP LV XVHG LQ UHPRWH
VHQVLQJ LPDJHU\ WR REWDLQ LQIRUPDWLRQ DERXW /DQG FRYHU
0DMRU FKDOOHQJHV DVVRFLDWHG ZLWK FODVVLILFDWLRQ DFFXUDF\
LQFOXGHGHVLJQSURFHGXUHRIFODVVLILHUFKRLFHRIWUDLQLQJVHWV
IURP GDWDVHW DQG LQIRUPDWLRQ FRQYH\HG WR WKH DOJRULWKP
6XSHULRULW\ RI GLIIHUHQW FODVVLILFDWLRQ DSSURDFKHV HPSOR\HG
GHSHQGV RQ VHOHFWHG GDWDVHW DQG WKH VWUDWHJ\ XVHG GXULQJ
GHVLJQLQJ SKDVH RI HDFK FODVVLILHU +RZHYHU LQ HQVHPEOH
DSSURDFK WKHUH LV QR GHILQLWH QXPEHU RI FODVVLILHUV WKDW
VKRXOG WDNH SDUW LQ GHFLVLRQ PDNLQJ 7KLV VWXG\ H[SORLWV
IHDWXUH VHOHFWLRQ WHFKQLTXH WR FUHDWH GLYHUVLW\ LQ HQVHPEOH
FODVVLILFDWLRQ 5HVXOWV REWDLQHG VKRZ WKDW IRU HQVHPEOH
DSSURDFKWKHUHLVQRVLJQLILFDQWEHQHILWLQKDYLQJPDQ\EDVH
FODVVLILHUV 7KH RXWFRPH VKRXOG UHYHDO KRZ WR GHVLJQ EHVW
HQVHPEOH XVLQJ IHDWXUH VHOHFWLRQ DSSURDFK IRU ODQG FRYHU
PDSSLQJ
.H\ZRUGV² 5HPRWH VHQVLQJ IHDWXUH VHOHFWLRQ
FODVVLILFDWLRQ/DQGFRYHUPDSSLQJ
,1752'8&7,21
,QSDWWHUQUHFRJQLWLRQODQGFRYHULQIRUPDWLRQLVNQRZQWREH
DQLPSRUWDQWGDWDFRPSRQHQWIRUYDULRXVDSSOLFDWLRQVDPRQJ
ZKLFK DUH HQYLURQPHQWDO DQDO\VLV JOREDO FKDQJH
LQYHVWLJDWLRQV DQG  LQ ODUJH VFDOH UHVRXUFH PDQDJHPHQW >@ 
>@5HPRWHVHQVLQJWHFKQRORJ\DSSOLFDWLRQKDVEHHQDPDMRU
WRRO IRU GHULYLQJ ODQG FRYHU LQIRUPDWLRQ 2QH PDMRU
DSSURDFK RI REWDLQLQJ ODQG FRYHU LQIRUPDWLRQ IURP UHPRWH
VHQVLQJ GDWD LV WKURXJK FODVVLILFDWLRQ 7KH FODVVLILFDWLRQ
DFFXUDF\SUREOHPUHODWHGWRODQGFRYHUPDSSLQJKDVEHHQDQ
LVVXH RI FRQFHUQ LQ SDWWHUQ UHFRJQLWLRQ 7KH SUREOHPV
LQFOXGH L WKH SURSRUWLRQ EHWZHHQ WKH QXPEHU RI WUDLQLQJ
VDPSOHV DQG WKH DYDLODEOH QXPEHU RI EDQGV DQG LL KLJK
FRUUHODWLRQEHWZHHQWKHWUDLQLQJSDWWHUQVIURPWKHVDPHDUHD
ZKLFKUHGXFHVWKHLQIRUPDWLRQ FRPPXQLFDWHGE\WKHVDPSOH
XQGHU FRQVLGHUDWLRQ WR WKH DOJRULWKP >@ ,Q PDFKLQH
OHDUQLQJ VHYHUDO FODVVLILFDWLRQ DOJRULWKPV KDYH EHHQ
GHYHORSHGIRUSURFHVVLQJUHPRWHO\VHQVHGGDWDVLQFHWKHILUVW
/DQGVDW LPDJH ZDV REWDLQHG LQ HDUO\ V >@ >@ 7KH
SXUSRVHLVWRILQGZD\VRILPSURYLQJFODVVLILFDWLRQDFFXUDF\
6XSHULRULW\ RI GLIIHUHQW DSSURDFKHV HPSOR\HG IRU
FODVVLILFDWLRQ GHSHQGV RQ WKH VHOHFWHG GDWD VHW DQG WKH
VWUDWHJ\ XVHG IRU WKH GHVLJQLQJ SKDVH RI HDFK FODVVLILHU
(QVHPEOHFODVVLILFDWLRQ LV DUHFHQWO\GHYHORSHGSDUDGLJPLQ
PDFKLQHOHDUQLQJDLPHGDWLPSURYLQJFODVVLILFDWLRQDFFXUDF\
RYHU VLQJOH DOJRULWKP ,Q HQVHPEOH FODVVLILFDWLRQ WKH
FRQVWLWXHQW FODVVLILHUV PXVW HUU LQ GLIIHUHQWO\ VR DV WR
LQWURGXFH GLYHUVLW\ LQWR WKH FODVVLILFDWLRQ V\VWHP $PRQJ
YDULRXVPHWKRGV XVHG WR PDNH DQ HQVHPEOHGLYHUVHIHDWXUH
VHOHFWLRQDSSURDFK KDV\LHOGHGFRQVLGHUDEOHJRRG UHVXOWVLQ
WHUPV RI FODVVLILFDWLRQ DFFXUDF\ >@ >@ >@ ,Q WKLV VWXG\
IHDWXUHVHOHFWLRQ WHFKQLTXHVDUHH[SORLWHGWRFUHDWHGLYHUVLW\
LQ HQVHPEOH FODVVLILFDWLRQ 7KH PHWKRGV RI FRPELQLQJ WKH
SUHGLFWLRQV RI WKH EDVH FODVVLILHUV LQWR DQ LQWHJUDWHG
HQVHPEOH FODVVLILFDWLRQ UHVXOW LQFOXGH PDMRULW\ YRWLQJ
ZHLJKWHGPDMRULW\YRWLQJFRQVHQVXVWKHRU\DQGVWDFNLQJ>@
>@>@)RUWKLVLQYHVWLJDWLRQPDMRULW\YRWLQJZDVDGRSWHG
DVWKH FRQVHQVXV UXOHIRU FRPELQLQJEDVH FODVVLILHUV8VLQJ
/DQGVDWLPDJHU\RI8JDQGDLQ(DVW$IULFDDQG$9,5,61:
,QGLDQD SLQHV WKH VWXG\IRFXVHV RQ LQYHVWLJDWLQJ WKH HIIHFWV
RI HQVHPEOH VL]H DQG IHDWXUH VHOHFWLRQ RQ FODVVLILFDWLRQ
DFFXUDF\RIODQGFRYHUPDSSLQJ
5(9,(:21(16(0%/(
(QVHPEOH WHFKQRORJ\ KDV EHHQ VXFFHVVIXOO\ XVHG LQ PDQ\
DSSOLFDWLRQV LQFOXGLQJ LPDJH FODVVLILFDWLRQ HJ FKDUDFWHU
UHFRJQLWLRQ KDQGZULWLQJ UHFRJQLWLRQ VSHHFK UHFRJQLWLRQ
DQG ILQJHUSULQWV UHFRJQLWLRQ >@ >@ >@ >@ ,QSDWWHUQ
UHFRJQLWLRQ DSSOLFDWLRQV RI HQVHPEOH PHWKRG KDV EHHQ
SURYHG WR EH DQ HIIHFWLYH WRRO IRU LQFUHDVLQJ FODVVLILFDWLRQ
DFFXUDF\ 7UDGLWLRQDOO\ GLYHUVLW\ LV LQWURGXFHG LQWR DQ
HQVHPEOH V\VWHP WKURXJK XVLQJ GLIIHUHQW FODVVLILHUV RU E\
WUDLQLQJ D FODVVLILHU ZLWKGLIIHUHQW WUDLQLQJ GDWDVHWV VXFK DV
ERRVWLQJ DQG EDJJLQJ >@ 7KH IHDWXUH EDQG VHOHFWLRQ
DSSURDFKDLPVDWVHOHFWLQJDVXEVHWRIWKHDYDLODEOHIHDWXUHV
2832978-1-4244-9566-5/10/$26.00 ©2010 IEEE IGARSS 2010
Authorized licensed use limited to: Makerere University Library. Downloaded on July 17,2021 at 20:47:00 UTC from IEEE Xplore. Restrictions apply.
7DEOH6XPPDU\RIWKHUHVXOWVREWDLQHGXVLQJ.DPSDODGDWDVHWZLWKWKHDSSOLFDWLRQRI6,
D
E
7DEOH5HVXOWVREWDLQHGXVLQJ%KDWWDFKDU\\D6,IHDWXUHVHOHFWLRQPHWKRG,QGLDQDGDWDVHWDQG
7DEOH(QVHPEOHFODVVLILFDWLRQXVLQJGLIIHUHQWVL]HVRIFODVVLILHUV
1RWH%& %DVHFODVVLILHU( (QVHPEOH6% 6LQJOH%HVWFODVVLILHU6, 6HSDUDELOLW\,QGH[
1) 1XPEHURI)HDWXUHVSHU%&%6 %DQGVVHOHFWHG' 'LYHUVLW\
6, %KDWWDFKDU\\D'LVWDQFH 'LYHUJHQFH
1)      
 %& %& %& %& %& %&
%6
           
           
           
           
           
(      
6%      
'
     
6, 7UDQVIRUPHG'LYHUJHQFH
)   
  %& %&
%      
     
     
     
     
(   
6%   
'
  
1XPEHURIIHDWXUHVSHU%DVH&ODVVLILHU
            
%&             
%&             
%&             
%&             
%&             
%&             
%&             
%&             
%&             
%&             
(             
6%             
1XPEHURI%&
 1XPEHURI%DQGVSHU(QVHPEOH
( ( ( ( ( ( (
       
       
       
       
       
       
       
       
2833
Authorized licensed use limited to: Makerere University Library. Downloaded on July 17,2021 at 20:47:00 UTC from IEEE Xplore. Restrictions apply.
WKDW  SURPRWH HIIHFWLYH GLYHUVLW\ EHWZHHQ WKH FODVVHV
XQGHU FRQVLGHUDWLRQ  KDYH EDQGV WKDW DUH DV PRVW
LQYDULDQW DV SRVVLEOH LQ WKH VSDWLDO ILHOG >@ >@ )HDWXUH
VHOHFWLRQ PHWKRGV KHOS LQ UHGXFLQJ WKH SURFHVVLQJ SHULRG
UHTXLUHG E\ WKH FODVVLILFDWLRQ SURFHVV DQG DOVR LPSURYH
FODVVLILFDWLRQDFFXUDF\ %\YDU\LQJWKH IHDWXUHVXEVHWVXVHG
WR FUHDWH WKH HQVHPEOH FODVVLILHU GLYHUVLW\ LV DVFHUWDLQHG
VLQFHWKHEDVHFODVVLILHUVWHQGWRHUULQGLIIHUHQWVXEVSDFHVRI
WKHLQVWDQFHVSDFH>@
9DULRXV PHWKRGV XVHG IRU VHOHFWLQJ IHDWXUHV IRU HQVHPEOH
V\VWHPV LQFOXGH H[KDXVWLYH VHDUFK UDQGRP VHOHFWLRQ RI
IHDWXUH VXEVHWV >@ DQG JHQHWLF DOJRULWKPV >@ )RU WKH
SXUSRVH RI WKLV VWXG\ ZH H[SORUHG IHDWXUH VHOHFWLRQ
DSSURDFK RQ UHPRWHO\ VHQVHG LPDJHU\ XVLQJ %KDWWDUFKDUU\D
GLVWDQFH 7UDQVIRUPHG GLYHUJHQFH DQG 'LYHUJHQFH
VHSDUDELOLW\ LQGH[ WR FUHDWH GLYHUVLW\ LQ WKH HQVHPEOH
PHWKRG 7KH JHQHUDWHG IHDWXUH VXEVHWV IURP WKH GDWDVHWV
ZHUHXVHGWRWUDLQDQGWHVWWKHEDVHFODVVLILHUV
0(7+2'2/2*<
7ZR GLIIHUHQW UHPRWHO\ VHQVHG GDWDVHWV ZHUH XVHG IRU WKLV
LQYHVWLJDWLRQ7KH ILUVW GDWDVHW LQFOXGHGD /DQGVDW LPDJH RI
.DPSDODWKHFDSLWDORI8JDQGD(DVW$IULFD7KHGDWDVHWZDV
JHQHUDWHGLQZLWKFROXPQDQGURZ 7KHLPDJH
FRQWDLQV ILYH ODQG FRYHU FODVVHV ZKLFK DUHZDWHU EXLOW XS
DUHDV WKLFN VZDPSV OLJKW VZDPSV DQG RWKHU YHJHWDWLRQV
7KUHH VHSDUDELOLW\ LQGLFHV ZHUH XVHG WR VHOHFW WKH IHDWXUHV
IRU FUHDWLQJ WKH EDVH FODVVLILHUV 7KH LQGLFHV DUH
%KDWWDFKDU\\D GLVWDQFH 'LYHUJHQFH DQG 7UDQVIRUPHG
GLYHUJHQFH)RUHDFKVHSDUDELOLW\LQGH[ILYHEDVHFODVVLILHUV
ZHUHJHQHUDWHGXVLQJWKHIHDWXUHVHOHFWHGIRUWKHHQVHPEOHV
7KH VHFRQG GDWDVHW LQFOXGHG WKH ,QGLDQ SLQHV VLWHLQ1RUWK
ZHVWHUQ ,QGLDQD K\SHUVSHFWUDO GDWD IURP $LUERUQH
9LVLEOH,QIUDUHG ,PDJLQJ6SHFWURPHWHU $9,5,6 >@ 7KH
GDWDVHWGHVLJQDWHG$9&FRQWDLQVEDQGVRXWRIZKLFK
 EDQGV ZHUH GLVFDUGHG EHFDXVH WKH\ ZHUH DIIHFWHG E\
DWPRVSKHULFFRQGLWLRQVEDQGVRXWRIWKHUHPDLQLQJ
ZHUHXVHGIRUWKLVLQYHVWLJDWLRQ7KHGDWDVHWLVDFFRPSDQLHG
ZLWK WKH JURXQG WUXWK LQIRUPDWLRQ KDYLQJ VL[WHHQ ODEHOHG
ODQG FRYHU FODVVHV 7KH ODQG FRYHU FODVVHV DUH $OIDOID
EXLOGLQJJUDVVWUHH GULYHV FRUQPLQLPXP WLOO FRUQQRWLO
FRUQ JUDVVSDVWXUH JUDVVSDVWXUHPRZHG JUDVV WUHHV KD\
ZLQGURZHG RDWV VR\EHDQVPLQLPXP WLOO VR\EHDQQRWLOO
VR\EHDQFOHDQVR\EHDQFOHDQVWRQHVWHHOWRZHUVZKHDWDQG
ZRRGV %KDWWDFKDU\\D GLVWDQFH VHSDUDELOLW\ LQGH[ ZDV XVHG
IRUH[KDXVWLYHIHDWXUHVHOHFWLRQVHDUFK
5HVHDUFKGHVLJQ
7KHGHVLJQ LQYROYHGLQFUHPHQWDOO\LQFUHDVLQJWKHQXPEHURI
EDVH FODVVLILHUV RQ RQH KDQG DQG RQ WKH RWKHU KDQG
LQFUHDVLQJ WKH QXPEHU RI EDQGV SHU EDVH FODVVLILHU %DVH
FODVVLILHUVIRUH[SHULPHQWZHUHLQFUHDVHGIURPWKUHHWR
WHQ  SHU HQVHPEOH 7KH QXPEHU RI EDQGV SHU EDVH
FODVVLILHU ZDV LQFUHPHQWDOO\ LQFUHDVHG LQ WKH IROORZLQJ
VHTXHQFHXSWRWKHQDQG
6XSSRUW 9HFWRU PDFKLQHV 690 ZKLFK KDV EHHQ IRXQG
FRPSHWLWLYH ZLWK WKH EHVW H[LVWLQJ PDFKLQH OHDUQLQJ
DOJRULWKPVLQ FODVVLI\LQJUHPRWHO\VHQVHG GDWDVHWV ZDVXVHG
WR FUHDWH WKH HQVHPEOHV 7KH 690 LV D VXSHUYLVHG
FODVVLILFDWLRQ DSSURDFK ZKLFK XVHV RSWLPL]DWLRQ DOJRULWKPV
WRILQG WKH RSWLPDO GHFLVLRQ ERXQGDULHVEHWZHHQ ODQG FRYHU
FODVVHVDQGRQO\VPDOO WUDLQLQJVDPSOHVDUHQHHGHG>@>@
5HVHDUFKHUV HPSOR\LQJ 690 IRU SDWWHUQ UHFRJQLWLRQ KDYH
FRQFOXGHG WKDW 690 SURYLGHV VLJQLILFDQW DGYDQWDJHV LQ
DFFXUDF\ VLPSOLFLW\ DQG UREXVWQHVV >@ %DVHG RQ WKH
GLVFRYHUHGDGYDQWDJHVLQWKHDSSOLFDWLRQRI690 *DXVVLDQ
VXSSRUWYHFWRUPDFKLQHZDVXVHGIRUWKHVWXG\
5(68/76$1'',6&866,21
7KH UHVXOWV REWDLQHG IURP WKH EDVH FODVVLILHUV DQG WKH
HQVHPEOHVZHUHFDOFXODWHGLQWHUPVRI.DSSD&RHIILFLHQWRI
$JUHHPHQW> @ E\FRPSDULQJ WKH JHQHUDWHG PDSV ZLWK WKH
JURXQG WUXWK G DWD 7DEOHV D QG  VKRZ WKH UHVXOWV REWDLQHG 
XVLQJ.DPSDODDQG,QGLDQDGDWDVHWVUHVSHFWLYHO\7KHUHVXOWV
REWDLQHG IURP WKH 7DEOHV LQ DOO FDVHV VKRZ WKDW
FODVVLILFDWLRQ DFFXUDF\ LQFUHDVHG DV WKH QXPEHU RIIHDWXUHV
LQWKH HQVHPEOHLQFUHDVH7KHUHVXOWV LQ7DEOH DOVRUHIOHFW
WKH PHDVXUH RI GLYHUVLW\ SHU HQVHPEOH DFFRUGLQJ WR WKH
GHJUHHRI DJUHHPHQW,WLV REVHUYHGIURPWKHUHVXOWVWKDWWKH
HQVHPEOH UHVXOW LQ HDFK LQVWDQFH LV EHWWHU WKDQ PDMRULW\ RI
WKH EDVH FODVVLILHU ¶V SUHGLFWLRQV WKRXJK WKH EHVW SUHGLFW LRQ
REWDLQHGIURPWKHEDVHFODVVLILHUFRQVWLWXWLQJHDFK HQVHPEOH
LV EHWWHU WKDQ WKH HQVHPEOH UHVXOW 7KH GLYHUVLW\ LQ WKH
SUHGLFWLRQV RI WKH EDVHFODVVLILHUV FRQVWLWXWLQJWKHHQVHPEOH
UHGXFHVWKHULVN RI PDNLQJ D SDUWLFXODU SRRUVHOHFWLRQ >@
>@ 7DEOH  VKRZV WKDW WKHUH DUH LQFUHDVHV LQ DFFXUDF\
UHVXOWV REWDLQHG EHWZHHQ ( DQG ( WKHQ ( DQG (
$FFXUDF\ UHVXOWV REWDLQHG EHWZHHQ ( DQG ( DUH
LQVLJQLILFDQW
7R IXUWKHU DXWKHQWLFDWH WKH UHVXOWV ELQRPLDO WHVW RI
VLJQLILFDQFH%76ZDVFRQGXFWHGRQWKHHQVHPEOHUHVXOWVWR
JLYH D EHWWHU DSSUHFLDWLRQ RI WKH GLIIHUHQFH EHWZHHQ WKH
HQVHPEOHV 7KLV LV WR DVFHUWDLQ WKH SDLU ZLVH GLIIHUHQFH
EHWZHHQWZRHQVHPEOHFODVVLILFDWLRQVVLGHVWHVW%76ZDV
FDUULHGRXWDWGHJUHHRIFRQILGHQFH7KHQXOOK\SRWKHVLV
K
R
VWDWLQJWKDWWKHUHLVQRVLJQLILFDQWGLIIHUHQFHEHWZHHQWKH
HQVHPEOHUHVXOWVLVKHUHE\UHMHFWHG LI µ]µ! >@7KH
GLIIHUHQW VHDUFK PHWKRGV XVHG LQGLFDWHV WKDW FODVVLILFDWLRQ
DFFXUDF\LQFUHDVHV DV WKHQXPEHURI IHDWXUHVPDNLQJXS WKH
HQVHPEOH LQFUHDVHV DQG WKDW WKHUH LV QR VLJQLILFDQW JDLQ LQ
KDYLQJPDQ\EDVHFODVVLILHUV
2834
Authorized licensed use limited to: Makerere University Library. Downloaded on July 17,2021 at 20:47:00 UTC from IEEE Xplore. Restrictions apply.
&21&/86,21
7KLV SDSHU KDV UHSRUWHG UHVXOWVREWDLQHG IURP WKH VWXG\RQ
HIIHFW RI HQVHPEOH RQ FODVVLILFDWLRQ DFFXUDF\ DV DSSOLHG WR
ODQGFRYHU PDSSLQJ7KH LQYHVWLJDWLRQZDVFDUULHGRXWXVLQJ
ODQGVDW DQG K\SHUVSHFWUDO GDWDVHWV ZLWK VXSSRUW YHFWRU
PDFKLQHVDVWKHEDVHFODVVLILHUV7KHVWXG\UHYHDOWKDW
LLQ HDFK LQVWDQFH HQVHPEOH UHVXOWV DUH EHWWHU WKDQ
PDMRULW\RIWKHEDVHFODVVLILHU¶VSUHGLFWLRQV
LLFODVVLILFDWLRQ DFFXUDF\ LQFUHDVH DV WKH QXPEHU RI
IHDWXUHVLQWKHEDVHFODVVLILHULQFUHDVHV
LLLWKHUHLV QR VLJQLILFDQW EHQHILW LQ KDYLQJ PDQ\EDVH
FODVVLILHUV 7KUHH EDVH FODVVLILHUVDUH VXIILFLHQW IRU
HQVHPEOHFODVVLILFDWLRQ
5()(5(1&(6
>@&+XDQJ /6 'DYLV DQG -5* 7RZQVHQG  ³$Q
DVVHVVPHQW RI VXSSRUW YHFWRU PDFKLQHV IRU ODQG FRYHU
FODVVLILFDWLRQ´ ,QWHUQDWLRQDO -RXUQDO RI 5HPRWH 6HQVLQJ
YROQRSS
>@%06WHHO DQG '$ 3DWWHUVRQ ³/DQG FRYHU PDSSLQJ
XVLQJ FRPELQDWLRQ DQG HQVHPEOH FODVVLILHUV´ &RPSXWLQJ
6FLHQFHDQG6WDWLVWLFVYROSS±
>@/%UX]]RQH³$QRYHODSSURDFKWRWKHVHOHFWLRQRI
VSDWLDOO\ LQYDULDQW IHDWXUHV IRU WKH FODVVLILFDWLRQ RI
K\SHUVSHFWUDO LPDJHV ZLWK LPSURYHG JHQHUDOL]DWLRQ
FDSDELOLW\´ ,(((7UDQVDFWLRQV RQ *HRVFLHQFH  5HPRWH
6HQVLQJYROQRSS±6HSW
>@)*+DOO-57RZQVHQGDQG(7(QJPDQ³6WDWXVRI
UHPRWH VHQVLQJ DOJRULWKPV IRU HVWLPDWLRQ RI ODQG VXUIDFH
VWDWHSDUDPHWHUV´ 5HPRWH 6HQVLQJ RI (QYLURQPHQW YRO
SS±
>@5 $UFKLEDOG DQG * )DQQ ³)HDWXUH VHOHFWLRQ DQG
FODVVLILFDWLRQ RI +\SHUVSHFWUDO LPDJHV ZLWK VXSSRUW YHFWRU
PDFKLQHV´ ,((( *HRVFLHQFH  5HPRWH 6HQVLQJ /HWWHUV
YROQR
>@5+XEHUDQG/9'XWUD³&ODVVLILHUFRPELQDWLRQDQG
IHDWXUH VHOHFWLRQ IRU ODQGFRYHU PDSSLQJ IURP KLJK
UHVROXWLRQ DLUERUQH GXDOEDQG6$5GDWD´ YRO SDUWSS
2UODQGR86$
>@' 2SWL] ³)HDWXUH VHOHFWLRQ IRU (QVHPEOHV´,Q
3URFHHGLQJV RI WKH 
WK
 1DWLRQDO &RQIHUHQFH RQ
$UWLILFLDO,QWHOOLJHQFH $$$, 2UODQGR)ORULGD 86$SS
±
>@6 7XO\DNRY 6 -DHJHU DQG 9 *RYLQGDUDMX DQG '
'RHUPDQQ ³5HYLHZ RI FODVVLILHU FRPELQDWLRQ PHWKRGV´
6WXGLHV LQ &RPSXWDWLRQDO ,QWHOOLJHQFH YRO  SS  ±

>@5 3ROLNDU ³(QVHPEOH EDVHG V\VWHPV LQ GHFLVLRQ
PDNLQJ´ ,((( &LUFXLWV DQG 6\VWHPV 0DJD]LQHWKLUG
TXDUWHUSS±
>@- .LWWOHU 0 +DWHI 53 'XLQ DQG - 0DWHV ³2Q
FRPELQLQJ FODVVLILHUV´ ,((( 7UDQVDFWLRQV RQ 3DWWHUQ
$QDO\VLV DQG 0DFKLQH ,QWHOOLJHQFH YRO  QR SS

>@/ ;X $ .U] \]DN&< 6XHQ ³0HWKRGV R I FRPELQLQJ
PXOWLSOH FODVVLILHUV DQG WKHLU DSSOLFDWLRQV WR KDQGZULWLQJ
UHFRJQLWLRQ´ ,((( 7UDQVDFWLRQV RQ 6\VWHP 0DQ DQG
&\EHUQHWLFVYROQRSS±
>@/ %UHLPDQ ³%DJJLQJ SUHGLFWRUV´ 0DFKLQH OHDUQLQJ
YROQRSS±
>@$ 7V\PEDO 0 3HFKHQL]NL\ DQG 3 &XQQXQJKDP
³'LYHUVLW\ LQ VHDUFK VWUDWHJLHV IRU HQVHPEOH IHDWXUH
VHOHFWLRQ´ ,QIRUPDWLRQ )XVLRQYRO  QR SS  ± 

>@7. +R ³7KH UDQGRP VXEVSDFH PHWKRG IRU
FRQVWUXFWLQJ GHFLVLRQ IRUHVWV´,(((7UDQVFDWLRQRI 3DWWHUQ
$QDO\VLV DQG 0DFKLQH,QWHOOLJHQFH YRO QR SS ±

>@$9,5,61:,QGLDQD¶V,QGLDQ3LQHVGDWDVHW
>2QOLQH@IWSIWSHFQSXUGXHHGXELHKO0XOWLVSHF$9&
2ULJLQDOILOHDQG
IWSIWSHFQSXUGXHHGXELHKO0XOWLVSHF7K\)LOHV]LS
JURXQGWUXWK
>@* &DPSV9DOOV DQG / %UX]]RQH ³.HUQHO EDVHG
PHWKRGV IRU K\SHUVSHFWUDO LPDJH FODVVLILFDWLRQ´ ,(((
7UDQVDFWLRQ*HRVFLHQFH 5HPRWH 6HQVLQJ YRO QR
SS±-XQH
>@*+5RVHQILHOGDQG.)LW]SDWULFN/LQV³$FRHIILFLHQW
RI DJUHHPHQW DV D PHDVXUH RI WKHPDWLF FODVVLILFDWLRQ
DFFXUDF\´ 3KRWRJUDPPHWULF (QJLQHHULQJ DQG 5HPRWH
6HQVLQJYROSS

2835
Authorized licensed use limited to: Makerere University Library. Downloaded on July 17,2021 at 20:47:00 UTC from IEEE Xplore. Restrictions apply.
... Furthermore, it is worth noting that land cover classification is one of the most important applications of remote sensing images. Some studies indicate significant improvements in the field of combination of classifiers, as highlighted by Abe, Gidudu, and Marwal (2010), with use feature selection techniques to create diversity in ensemble classification. Tinoco et al. (2013), for example, showed the efficiency in using an ensemble of classifiers concerning hyperspectral land cover analysis. ...
Pruning Optimum-Path Forest Ensembles Using Metaheuristic Optimization for Land-Cover Classification
Article
  • Apr 2017
Machine learning techniques have been actively pursued in the last years, mainly due to the increasing number of applications that make use of some sort of intelligent mechanism for decision-making processes. In this context, we shall highlight ensemble pruning strategies, which provide heuristics to select from a collection of classifiers the ones that can really improve recognition rates and provide efficiency by reducing the ensemble size prior to combining the model. In this paper, we present and validate an ensemble pruning approach for Optimum-Path Forest (OPF) classifiers based on metaheuristic optimization over general-purpose datasets to validate the effectiveness and efficiency of the proposed approach using distinct configurations in real and synthetic benchmark datasets, and thereafter, we apply the proposed approach in remote sensing images to investigate the behavior of the OPF classifier using pruning strategies. The image datasets were obtained from CBERS-2B, Landsat-5 TM, Ikonos-2 MS, and GeoEye sensors, covering some areas of Brazil, and well-known Indian Pines. In this work, we evaluate five different optimization algorithms for ensemble pruning , including that Particle Swarm Optimization, Harmony Search, Cuckoo Search and Firefly Algorithm. In addition, we performed an empirical comparison between Support Vector Machine (SVM) and OPF using the strategy of ensemble pruning. Experimental results showed the effectiveness and efficiency of ensemble pruning using OPF-based classification, especially concerning ensemble pruning using Harmony Search, which shows to be effective without degrading the performance when applied to large datasets, as well as a good data generalization.
View
... Research has shown that ensemble generates better classification accuracy results than the individual classifier making up the ensemble [4,14,15]. ...
Classification of Hyperspectral Images Using Machine Learning Methods
Article
  • Sep 2014
Mixed pixels problem has significant effects on the application of remote sensing images. Spectral unmixing analysis has been extensively used to solve mixed pixels in hyperspectral images. This is based on the knowledge of a set of unidentified endmembers. This study used pixel purity index to extract endmembers from hyperspectral dataset of Washington DC mall. Generalized reduced gradient (GRG) a mathematical optimization method is used to estimate fractional abundances (FA) in the dataset. WEKA data mining tool is chosen to develop ensemble and non-ensemble classifiers using the set of the FA. Random forest (RF) and bagging represent ensemble methods while neural networks and C4.5 represent non-ensemble models for land cover classification (LCC). Experimental comparison between the classifiers shows that RF outperforms all other classifiers. The study resolves the problem associated with LCC by using GRG algorithm with supervised classifiers to improve overall classification accuracy. The accuracy comparison of the learners is important for decision makers in order to consider tradeoffs in accuracy and complexity of methods.
View
Artificial Intelligence in Politics
Chapter
  • Aug 2023
Artificial intelligence (AI) has recently emerged as a potent force for change, touching every business and every aspect of modern life. AI’s immense powers are drastically transforming international politics, decision-making and security. Because AI can swiftly analyze massive volumes of data and discover patterns, it opens up new opportunities for more effective policy creation, sophisticated diplomatic conversations, and preemptive threat detection. However, bringing AI into international politics raises challenges and moral concerns like accountability, algorithmic bias, and data privacy. Therefore, understanding the implications of AI in international politics to develop a collaborative mindset is critical as governments navigate this new frontier to achieve responsible, equitable, and secure outcomes for all stakeholders.KeywordsArtificial intelligenceNeural networksEvolutionary programmingDeep learningFuzzy logic
View
Hyperspectral data classification based on local polynomial approximation
Article
  • Jan 2015
Processing and classification of hyperspectral data into various class labels have several challenges due to image objects entrenchment in a single pixel and large data size that needs serious computation and huge memory. This work presence local polynomial approximation (LPA) method to process Washington DC Mall hyperspectral data set for image classification. The data generated from LPA are then classified using neural network (NN), support vector machine (SVM) and random forest (RF). The classification procedures are implemented in Waikato Environment for Knowledge Analysis (WEKA). To evaluate the results, the performer of the classifiers per class label are presented in metrics and the overall classification results are presented in tabular form. The Friedman statistical test is carried out on the classification results to establish the performance of each classifier on the processed data. The LPA method is evaluated on the datasets using the different classifiers to demonstrate its efficacy.
View
Hyperspectral Image Classification based on NMF Features Selection Method
Article
  • Nov 2013
  • Proceedings of SPIE
Hyperspectral instruments are capable of collecting hundreds of images corresponding to wavelength channels for the same area on the earth surface. Due to the huge number of features (bands) in hyperspectral imagery, land cover classification procedures are computationally expensive and pose a problem known as the curse of dimensionality. In addition, higher correlation among contiguous bands increases the redundancy within the bands. Hence, dimension reduction of hyperspectral data is very crucial so as to obtain good classification accuracy results. This paper presents a new feature selection technique. Non-negative Matrix Factorization (NMF) algorithm is proposed to obtain reduced relevant features in the input domain of each class label. This aimed to reduce classification error and dimensionality of classification challenges. Indiana pines of the Northwest Indiana dataset is used to evaluate the performance of the proposed method through experiments of features selection and classification. The Waikato Environment for Knowledge Analysis (WEKA) data mining framework is selected as a tool to implement the classification using Support Vector Machines and Neural Network. The selected features subsets are subjected to land cover classification to investigate the performance of the classifiers and how the features size affects classification accuracy. Results obtained shows that performances of the classifiers are significant. The study makes a positive contribution to the problems of hyperspectral imagery by exploring NMF, SVMs and NN to improve classification accuracy. The performances of the classifiers are valuable for decision maker to consider tradeoffs in method accuracy versus method complexity.
View
Ensemble Classifier Using GRG Algorithm for Land Cover Classification
Article
  • Nov 2013
  • Proceedings of SPIE
Image processing is of great value because it enables satellite images to be translated into useful information. The preprocessing of remotely sensed images before features extraction is important to remove noise and improve the ability to interpret image data more accurately. All images should appear as if they were acquired from the same sensor at the end of image preprocessing. A major challenge associated with hyperspectral imagery in remote sensing analysis is the mixed pixels which are due to huge dimension nature of the data. This study makes a positive contribution to the problem of land cover classification by exploring Generalized Reduced Gradient (GRG) algorithm on hyperspectral datasets by using Washington DC mall and Indiana pines test site of Northwestern Indiana, USA as study sites. The algorithm was used to estimate the fractional abundance in the datasets for land cover classification. Ensemble classifiers such as random forest, bagging and support vector machines were implemented in Waikato Environment for knowledge Analysis (WEKA) to carry out the classification procedures. Experimental results show that random forest ensemble outperformed the other ensemble methods. The comparison of the classifiers is crucial for a decision maker to consider compromises in accuracy technique against complexity technique.
View
View
An assessment of support vector machines for land cover classié cation
Article
Full-text available
  • Feb 2002
The support vector machine (SVM) is a group of theoretically superior machine learning algorithms. It was found competitive with the best available machine learning algorithms in classifying high-dimensional data sets. This paper gives an introduction to the theoretical development of the SVM and an experi- mental evaluation of its accuracy, stability and training speed in deriving land cover classié cations from satellite images. The SVM was compared to three other popular classié ers, including the maximum likelihood classié er (MLC), neural network classié ers (NNC) and decision tree classié ers (DTC). The impacts of kernel coné guration on the performance of the SVM and of the selection of training data and input variables on the four classié ers were also evaluated in this experiment.
View
Review of Classifier Combination Methods
Chapter
Full-text available
  • Dec 2007
Classifier combination methods have proved to be an effective tool to increase the performance of pattern recognition applications. In this chapter we review and categorize major advancements in this field. Despite a significant number of publications describing successful classifier combination implementations, the theoretical basis is still missing and achieved improvements are inconsistent. By introducing different categories of classifier combinations in this review we attempt to put forward more specific directions for future theoretical research.We also introduce a retraining effect and effects of locality based training as important properties of classifier combinations. Such effects have significant influence on the performance of combinations, and their study is necessary for complete theoretical understanding of combination algorithms.
View
A Novel Approach to the Selection of Spatially Invariant Features for the Classification of Hyperspectral Images With Improved Generalization Capability
Article
Full-text available
  • Oct 2009
This paper presents a novel approach to feature selection for the classification of hyperspectral images. The proposed approach aims at selecting a subset of the original set of features that exhibits at the same time high capability to discriminate among the considered classes and high invariance in the spatial domain of the investigated scene. This approach results in a more robust classification system with improved generalization properties with respect to standard feature-selection methods. The feature selection is accomplished by defining a multiobjective criterion function made up of two terms: (1) a term that measures the class separability and (2) a term that evaluates the spatial invariance of the selected features. In order to assess the spatial invariance of the feature subset, we propose both a supervised method (which assumes that training samples acquired in two or more spatially disjoint areas are available) and a semisupervised method (which requires only a standard training set acquired in a single area of the scene and takes advantage of unlabeled samples selected in portions of the scene spatially disjoint from the training set). The choice for the supervised or semisupervised method depends on the available reference data. The multiobjective problem is solved by an evolutionary algorithm that estimates the set of Pareto-optimal solutions. Experiments carried out on a hyperspectral image acquired by the Hyperion sensor on a complex area confirmed the effectiveness of the proposed approach.
View
View
Land Cover Mapping Using Combination and Ensemble Classifiers
Article
  • Jan 2002
In recent years, large scale land cover maps constructed from remotely sensed data have become important information sources for resource management. Classifiers are commonly used to predict land cover for unsampled map units; hence, they play a key role in map construction. Achieving adequate classifier accuracy is often problematic for highly variable and difficult-to-sample landscapes. This article investigates a variety of methods for improving accuracy based on 1) combining a few different classifiers, and 2) creating ensembles of many classifiers. In addition, we derive an analytic expression for the exact bagging -nearest neighbor classifier.
View
A coefTcient of agreement as a measure of thematic classiTcation accuracy
Article
  • Jan 1986
  • PHOTOGRAMM ENG REM S
The classification error matrix typically contains tabulated results of accuracy evaluation for a thematic classification, such as a land-use and land-cover map. Diagonal elements of the matrix represent counts correct. The usual designation of classification accuracy has been total percent correct. Nondiagonal elements of the matrix have usually been neglected. A coefficient of agreement is determined for the interpreted map as a whole, and individually for each interpreted category. These coefficients utilize all cell values in the matrix.-from Authors
View
Bagging Predictors
Article
  • Aug 1996
Bagging predictors is a method for generating multiple versions of a predictor and using these to get an aggregated predictor. The aggregation averages over the versions when predicting a numerical outcome and does a plurality vote when predicting a class. The multiple versions are formed by making bootstrap replicates of the learning set and using these as new learning sets. Tests on real and simulated data sets using classification and regression trees and subset selection in linear regression show that bagging can give substantial gains in accuracy. The vital element is the instability of the prediction method. If perturbing the learning set can cause significant changes in the predictor constructed, then bagging can improve accuracy.
View
Diversity in search strategies for ensemble feature selection
Article
  • Mar 2005
  • INFORM FUSION
Ensembles of learnt models constitute one of the main current directions in machine learning and data mining. Ensembles allow us to achieve higher accuracy, which is often not achievable with single models. It was shown theoretically and experimentally that in order for an ensemble to be effective, it should consist of base classifiers that have diversity in their predictions. One technique, which proved to be effective for constructing an ensemble of diverse base classifiers, is the use of different feature subsets, or so-called ensemble feature selection. Many ensemble feature selection strategies incorporate diversity as an objective in the search for the best collection of feature subsets. A number of ways are known to quantify diversity in ensembles of classifiers, and little research has been done about their appropriateness to ensemble feature selection. In this paper, we compare five measures of diversity with regard to their possible use in ensemble feature selection. We conduct experiments on 21 data sets from the UCI machine learning repository, comparing the ensemble accuracy and other characteristics for the ensembles built with ensemble feature selection based on the considered measures of diversity. We consider four search strategies for ensemble feature selection together with the simple random subspacing: genetic search, hill-climbing, and ensemble forward and backward sequential selection. In the experiments, we show that, in some cases, the ensemble feature selection process can be sensitive to the choice of the diversity measure, and that the question of the superiority of a particular measure depends on the context of the use of diversity and on the data being processed. In many cases and on average, the plain disagreement measure is the best. Genetic search, kappa, and dynamic voting with selection form the best combination of a search strategy, diversity measure and integration method.
View
Status of remote sensing algorithms for estimation of land surface state parameters
Article
  • Jan 1995
  • REMOTE SENS ENVIRON
Ecosystem process, biosphere-atmosphere transfer, and carbon exchange models all require parameterization of the land surface, including land vegetation cover and soil moisture. Although not yet a demonstrated global capability, the most feasible method for obtaining these parameters and updating them periodically, is satellite remote sensing. In this paper we will summarize our understanding of the desired land surface parameters, including soil moisture, and provide an assessment of the state of the art of surface state remote sensing algorithms to infer those parameters on a global basis.First, we will consider a) modeling requirements for land cover parameters, including vegetation community composition and biophysical parameters, for example, leaf area index (LAI), biomass density, fraction of photosynthetically active radiation (Fear) absorbed by the vegetated land surface, and b) modeling requirements for soil moisture.We will then review the status of remote sensing algorithms for obtaining these parameters and examine a number of issues involved in the global implementation and testing of these algorithms. Finally, we will look at future needs to make global mapping of land cover parameters a reality.
View