913
PHOTOGRAMMETRIC45075820912867111141183222254291333374402433465478515
ENGINEERING97975820912769109122164191218250263305345
&1338758209136
REMOTE138875820913259101145174201
SENSING16037582091265395121134176215
March502375930785592121153191
2014523875930784181120155
261563075910803570100
Nested7535860359157286401488616747
Regression1561586035915928441850263274786192510551188
Based28105860359157289403533664
Optimal35345860359155286372437636768832
Selection442658603591412703354645886747388681001
(NRBOS)739896035971228387544699840911
of17118960359129207
Rational19788960359160291376440570703834898
Polynomial2937896035914627333744557870890797211031167
Coefficients41658960359152281411488556624748813942107511601275
Long2332149101144593143193
Tengfei,2547149101144488137187216263287310
Jiao2879149101143559107156
Weili,30571491011479124148172196219
and3299149101144898146
He34681491011458106
Guojin35971491011458106155179203252
Abstract450177101515197137167196241284314
Although450190201076897127182232287332387
the862190201073085130
rational1017190201073585115143193248298326
function1368190201073085140185215244294349
model17421902010780130185230258
(RFM)2025190201073020551919086489115522101902010730
is2265190201072868
widely2358190201076997152197225272
applied26551902010750105160188217262317
in450200201072883
photogrammetry,5582002010755110160190240285320370450530575605640687706
the1289200201072984129
application14442002010750105160188216261311341370420475
of1944200201075080
terrain-dependent2049200201073075110145195223278308363408463508563618663718748
RFM282220190864891155
is450210201072868
limited543210201072957137165195240295
because863210201075095140190245285330
of1218210201075080
the1323210201073085130
requirement1478210201073580130185214249294374419474504
for2007210201073080115
numerous21472102010755110190235270320375415
ground2587210201074580130185240295
control450220201074595150180215265293
points7682202010755105134189219259
(GCPs)10522202010730108222190865210014417612582202010730
and13132202010750105160
the1498220201073085130
strong1653220201074070105155210255
correlation1933220201074595130165210238288318347397452
between2410220201075095125193238283338
the2773220201073085130
coefficients.450230201074595140170198225270298343398428468493
A9682302010769
new10622302010755100168
method,12552302010780125155210260315340
NRBOS,162023190865610415120725118712302010725
based19212302010750100140185240
on21862302010750105
nested23162302010755100140170215270
regression2611230201073580125160205245450240201074068118173
was6482402010769119159
proposed832240201075590140195245285330385
to1242240201073080
select1347240201074085113158203233
the1605240201073085130
optimal17602402010750105135163243293322
RPCs210724190864892140172
automatically23042402010750105135185265315345373418468497525572
and4502502010750105160
to635250201073080
gain740250201074595123178
stable943250201074070120170199244
solutions1212250201074090118173203232282337376
of1614250201075080
terrain-dependent1719250201073075110145195223278308363408463508563618663718748
RFMusing2492251908648911552672250201075595123178223
a4502602010750
small5252602010740120170198227
amount7772602010750130180235290320
of1122260201075080
GCPs.122726190865210014417614032602010725
Different1453260201076593123153198233278333363
types1841260201073077132177217
of2083260201075080
images,21882602010728108158203248288313
including2526260201072984129157212267295350395
QuickBird,4502702010770125153198248307335370425450
SPOT5,9252719086448814419223211572702010725
Landsat-5,12072702010750100155210250300330360410435
andALOS,1667270201075010516018522719086559515119520472702010725
were20972702010768113148193
involved2315270201072984130180209256301356
in2696270201072883
the450280201073085130
tests.605280201073075115145185210
NRBOS840281908656104151207251
method11162802010780125155210260315
performed14562802010755100135165215250330375430
better1911280201075095125155200235
than2171280201073085135190
conventional2386280201074595150197242297327355405460510539
methods4502902010780125155210260315355
in830290201072883
estimating938290201074585115144224274304332387432
RPCs,13952919086489214017215672902010725
and16172902010750105160
even1802290201074592137192
provided2019290201075590140187215270315370
a24142902010750
reliable2489290201073580108137187237265310
solution450300201074090118173203232282337
when8123002010768123168223
less1060300201072873113153
than1238300201073085135190
39GCPs145330020107501001578301908652100144176
were17793002010768113148193
used.1997300201075595140195220
Additionally,22423002010769124179207237265315370420449477524542
the450310201073085130
test605310201073075115145
results775310201073580120175203233273
showed1073310201074095145213258313
that1411310201073085135165
the1601310201073085130
simplified1756310201074069149204232260288315360415
RPCs219631190864892140172
are2393310201075085130
almost2548310201075078158209248278
as450320201075090
accurate565320201075095140195230280310355
as945320201075090
the1060320201073085130
vendor-provided1215320201074792147202252287313368403454500529584629684
RPCs.19243219086489214017220963202010725
Consequently,21463202010760110165205250300355400455485513560578
in450330201072883
favorable558330201073080127177212262312341386
situations96933020107406898153203233261311366406
such1400330201074095140195
as1620330201075090
when17353302010769123169223
the1984330201072984129
orientation2138330201075186114159214244294324352402457
parameters26203302010755105450340201073585165210240285320360
of835340201075080
the940340201073085130
satellite1095340201074090120165193222250280325
are1445340201075085130
not16003402010755105135
available1760340201075097147175204254304332377
or2162340201075085
are2272340201075085130
not24273402010755105135
sufficiently2587340201074095125153180450350201074573118173203232278
accurate,753350201075095140195230280310355380
the1158350201073085130
proposed1313350201075590141195246285330385
method17233502010781126155210261315
has20643502010755105145
the2234350201073084129
potential23893502010754105135180234264293343371
to450360201073080
take555360201073080130175
the755360201073085130
place910360201075583133178223
of1158360201075180
the1264360201072984129
regular1418360201073680126180209259294
terrain-independent1737360201073075110145195223278308337392447492547592647702747802832
RFM.25943619086489115527483602010725
Introduction450386101512470100129174219265308337360405451
The4503992010760115160
rational635399201073580110140190245290320
function980399201073085140185215245295350
model13553992010780130185230260
(RFM)1640399201073016704009085489216018303992010730
with18853992010775105135190
7821003992010750100
rational2225399201073580110140190245290320
polynomial257039920107551051351852402904504092010780110155185
coefficients660409201074595140170197225270300345400430470
(RPCs)1155409201073011854109085489214017213574092010730
is1412409201073070
completely1507409201074595175230260305335380410460
a19924092010745
mathematical20624092010780125155210255335380410440485530560
model,26474092010780130185230260285
which4504192010775130160205260
approximately7354192010745100155190240290320400445475520550600
describes13604192010755100140185220250300345385
the1770419201073085130
imaging19254192010730110155200230285330
process2280419201075590140185230270310
in2615419201073085
photogrammetry27254192010755110160190240450429201074580125205285330360395445
and9204292010745100155
remote1100429201073580160210240285
sensing.1410429201074085140180210265310335
Terrain-independent1770429201076097132167212242297327357412467512567612667722767822852
RFM264743090854892160
constitutes2832429201074595150450439201074070100130185215260300
a7754392010745
comprehensive845439201074595175230265310365410465505535585630
reparameterization1500439201073580135180215260340385415460495525570615645675725780
of2305439201075080
the2410439201073085130
rigorous2565439201073565110160195245300340
sensor450449201074085140180230265
model,7404492010780130185230260285
and10504492010745100155
is1230449201073070
widely13254492010775105160205235285
applied16354492010745100155185215260315
in1975449201073085
high2085449201075585130185
resolution2295449201073580120170200255285315365420
image27404492010730110155200245
products.450459201075590140195250295325365390
Terrain-dependent865459201076097132167212242297327382427482527582637682737767
RFM,16574609085489216018174592010725
on18674592010750105
the1997459201073085130
other2152459201075080135180215
hand,23924592010755100155210235
is2652459201073070
hardly27474592010755100135190220270
used450469201075595140195
because670469201075095140185240280325
of1020469201075080
the1125469201073085130
requirement1280469201073580130185215250295375420475505
for1810469201073080115
numerous19504692010755110190235270320375415
observation23904692010750100140185220270315345375425480
data4504792010755100130175
(ground650479201073075110160215270325
control1000479201074595150180215265295
points13204792010755105135190220260
(GCPs)),160547920107301635480908552100144176181147920107306085
and19214792010745100155
the2101479201073085130
strong2256479201074070105155210255
correlation2536479201074595130165210240285315345395450
between450489201075095125200245290345
the820489201073085130
coefficients.975489201074595140170197225270300345400430470495
Actually,14954892010770115145200245275305355373
terrain-dependent1893489201073075110145190220275305360405460505560615660715745
RFM266349090854892160
provides450499201075590140190220275320360
a8354992010745
useful905499201075595140170225255
approach11854992010745100155190240285330385
to1595499201073080
rectify1700499201073580125155185215265
remotely1990499201073580160210240285315365
sensed2380499201074085140180225280
images26854992010730110155200245285
without4505092010775105135190240295325
knowing8005092010750105155230260315360
the1185509201073085130
position13405092010755105145175205235285340
and17055092010745100155
orientation1885509201075085115160215245290320350400455
information2365509201073085115165200280325355385435490
of450519201075080
specific555519201074095140185215242270315
sensor.895519201074085140180230265285
However,12055192010770120195240290335370390
as1620519201074585
7817305192010750100
RPCs185552090854892140172
of2052519201075080
the2157519201073085130
RFM231252090854892160
are2497519201074580125
strongly2647519201074070105155210255450529201073080
correlated,555529201074595130165210240285315360415440
stable,1020529201074070115165195240265
and13105292010745100155
precise1490529201075590135180210250295
solutions1810529201074090120175205235285340380
of2215529201075080
the2320529201073085130
RPCs247553090854892140172
are2672529201074580125
difficult450539201075585115142170215270300330
or805539201075085
even915539201074595140195
impossible11355392010730110165215255295325375405450
to1610539201073080
achieve1715539201074590145175220270315
(Lin2055539201073085115170
and22505392010745100155
Yuan,24305392010765115160215240
2008).26955392010750100150200230255
The4505492010760115160
objectives6355492010750100125170215245275325370410
of1070549201075080
this1175549201073085115155
research1355549201073580120165210245290345
are1725549201074580125
to1875549201073080
find1980549201072855110165
a21705492010745
robust2240549201073585135190230260
approach25255492010745100155190240285330385
to2935549201073080
estimate450559201074585115145225270300345
RPCs82056090854892140172
and10175592010745100155
to1197559201073080
make13025592010780125175220
use1547559201075595140
of1712559201075080
the1817559201073085130
terrain-dependent1972559201073075110145190220275305360405460505560615660715745
RFM.27425609085489216029025592010725
Many5705692010785130185235
studies830569201074070125180210255295
have11505692010755100150195
been1370569201075095140195
carried1590569201074590125160190235290
out19055692010750105135
on20655692010750105
the2195569201073085130
topic2350569201073080135165210
of2585569201075080
RFM269057090854892160
during4505792010755110145175230275
the750579201073085130
recent905579201073580125170225255
decades.11855792010755100145190245290330355
OGC1565580908556108156
(1999)1746579201073080130180230260
has20315792010755100140
normalized21965792010755105140220265295325370415470
the2691579201073085130
range450589201073580135180225
of700589201075080
the805589201073085130
image9605892010730110155200245
and12305892010745100155
object14105892010750100125170215245
space1680589201074095140185230
coordinates1935589201074595145180235265320365395440480
of2440589201075080
RFM254559090854892160
to2730589201073080
−128355892010755105
to450599201073080
+1,5555992010755105130
and7105992010745100155
effectively890599201074575105150195225255305350380430
enhanced13455992010745100155200255300345400
the1770599201073085130
condition1925599201074595150205235265294345400
number23505992010755110190240285320
of2695599201075080
the2800599201073085130
normal4506092010755105140220265295
equation770609201074595150195225255305360
matrix.11556092010780125155190220270295
Tao1475609201076097147
and16476092010745100155
Hu18276092010770125
(2000)1977609201073080130180230260
have22626092010755100150195
studied2482609201074070125180210255310
the2817609201073085130
RFM45062090854892160
comprehensively635619201074595175230265310365410465505535585630660710
and13706192010745100155
have15506192010755100150195
proposed1770619201075590140195245285330385
to2180619201073080
strengthen2285619201074070105150205250280335380435
the2745619201073085130
solution450629201074090120175205235285340
of815629201075080
RFMusing920630908548921601105629201075595125180225
Tikhonov1355629201076083133188238293343394
regularization1774629201073479124179209254289319364409439469519574
and23736292010745100155
the2553629201073085130
L-curve2708629201075585130185220270315
method.4506392010780125155210260315340
Yuan8156392010765115160215
and10556392010745100155
Lin1235639201075585140
(2008)1400639201073080130180230260
have16856392010755100150195
compared1905639201074595175230275310355410
the2340639201073085130
results2495639201073580120175205235275
of2795639201075080
several450649201074085135180215260290
methods7656492010780125155210260315355
for1145649201073080115
solving1285649201074090120170200255300
RPCs161065090854892140172
including1807649201073085130160215270300355400
ridge2232649201073565120165210
trace2467649201073065110155200
method,269264920107801251552104506592010750105130
L-curve605659201075585130185220270315
method,9456592010780125155210260315340
empirical13106592010745125180210245275320365395
formula1730659201073080115195250280325
method,20806592010780125155210260315340
and24456592010745100155
generalized2625659201074590145190225270300450669201073075120175
ridge650669201073565120165210
estimate885669201074585115145225270300345
method;12556692010780125155210260315340
they1620669201073085130180
have18256692010755100150195
verified2045669201075095130160187215260315
the2385669201073085130
validity2540669201075095125155210240270320
of2885669201075080
L-curve450679201075585130185220270315
method.7906792010780125155210260315340
In1155679201073085
addition,12656792010745100155185215245295350375
Levenberg-Marquardt16656792010755100150195250300345380425455540585620670725770805860890
method25806792010780125155210260315
(Tao3157181101073090127177
and33591811010745100155
Hu,35391811010770125150
2001)37141811010750100150200230
and39691811010745100155
singular4149181101074070125170225255300335
value4509181101075095125180225
decomposition47591811010755100145195275330380420450480510560615
method53991811010780125155210260315
have31571911010755100150195
been3377191101075095140195
applied35971911010745100155185215260315
to3937191101073080
solve4042191101074090120170215
RPCs428319270854892140172
(Fraser4480191101073085120165205250285
et4790191101074575
al.,48901911010750784968191101072550
2006).50431911010750100150200230255
Among53231911010770150200255300
all3158201101074575105
these3288201101073085130170215
methods,35282011010780125155210260315355380
the3933201101073085130
ridge4088201101073565120165210
estimation4323201101074585115145225270300330380435
method47832011010780125155210260315
(especially5123201101073075115170215260290335365395445
the5593201101073085130
L-curve3158211101075585130185220270315
method)34982111010780125155210260315345
is3868211101073070
the3963211101073085130
most41182111010780130170200
widely43432111010775105160205235285
used4653211101075595140195
approach.48732111010745100155190240285330385410
However,53082111010770120195240290335370390
there3158221101073085130165210
are3393221101074580125
still3543221101074070100130160
some3728221101074090170215
problems3968221101075590140190220265345385
in4378221101073085
solving4488221101074090120170200255300
RPCs481222270854892140172
using5009221101075595125180225
the5259221101073085130
existing54142211010745951251651953157231101073085130
methods.33122311010780125155210260315355380
For37172311010755105140
example,3882231101074595140220275305350375
ridge4282231101073565120165210
estimate4517231101074585115145225270300345
is4887231101073070
a49822311010745
biased5052231101075080125165210265
estimate,5342231101074585115145225270300345370
which31572411010775130160205260
requires3442241101073580130185215250295335
numerous38022411010755110190235270320375415
GCPs4242242708552100144176
to4443241101073080
solve4548241101074090120170215
RPCs.47882427085489214017249602411010725
Solving32772511010755105135185215270315
RPCs361825270854892140172
is3815251101073070
a39102511010745
problem3980251101075590140190220265345
of4350251101075080
multiple44552511010780135165195225280310355
regression4835251101073580125160205245285315365420
analysis.528025110107451001451752253158261101074070110135
The33182611010760115160
problem3503261101075590140190220265345
of3873261101075080
multicollinearity39782611010780135165195225270320350380410465510555590620650700
results4703261101073580120175205235275
in5003261101073085
an51132611010745100
ill-posed523826110107306090120175225265310365
normal31582711010755105140220265295
equation,3478271101074595150195225255305360385
and38882711010745100155
the4068271101073085130
ordinary4223271101075085140170225270305355
least4603271101073075120160190
squares4818271101074090145190225270310
(OLS)51532711010730518227270855610014453262711010730
estimation53812711010745851151452252703157281101073060110165
does33472811010755105150190
badly3562281101075095150180230
in3817281101073085
achieving3927281101074590145175220270300355400
a43522811010745
stable4422281101074070115165195240
and46872811010745100155
reliable4867281101073580110140185235265310
solution.5202281101074090120175205235285340365
In3157291101073085
order3267291101075085140185220
to3512291101073080
solve3617291101074090120170215
the3857291101073085130
problem4012291101075590140190220265345
of4382291101075080
multicollinearity,44872911010780135165195225270320350380410465510555590620650700718
“variable5230291101074595140175205250300330375
selection3157301101074085115160205235265315370
(Draper3552301101073095130175230275310
et3887301101074575
al.,39873011010750784066301101072550
1966)”41413011010750100150200230275
and44413011010745100155
“ridge4621301101074580110165210255
estimation4901301101074585115145225270300330380435
(Hoerl53613011010730100150195230260
and31573111010745100155
Kennard,33373111010765110165220265300355380
1970)”37423111010750100150200230275
are4042311101074580125
usually4192311101075595150195225255305
used4522311101075595140195
to4742311101073080
improve48473111010730110165200250300345
the5217311101073085130
OLS.537331270855610014455173111010725
Variable31573211010765103138168213263293338
selection3520321101074085115160205235265315370
can3915321101074590145
simplify4085321101074070150205235265295345
the4455321101073085130
original4610321101075085115160190245290320
model49553211010780130185230260
by52403211010750100
selecting5365321101073984114159204234264319364
a31573311010745
subset3227331101074095145185230260
of3512331101075080
variables3617331101075095130160205255285330370
from4012331101073065115195
the4232331101073085130
original4387331101075085115160190245290320
set4732331101074085115
of4872331101075080
variables4977331101075095130160205255285330370
which53723311010775130160205260
give3157341101074575125170
the3352341101073085130
most35073411010780130170200
significant3732341101074070115170200227255300345400430
response4187341101073580120175225280320365
to4577341101073080
the4682341101073085130
regression,4837341101073580125160205245285315365420445
and53073411010745100155
therefore54873411010730851301652103157351101073080115160
the3342351101073085130
multicollinearity34973511010780135165195225270320350380410465510555590620650700
of4222351101075080
the4327351101073085130
model44823511010780130185230260
can4767351101074590145
be4937351101075095
reduced5057351101073580135190235280335
after5417351101074575105150185
the3157361001073085130
variable3312361001075095130160205255285330
selection3667361001074085115160205235265315370
(Guyon4062361001073095150200250305
and43923610010745100155
Elisseeff,4572361001075585115155195240285315345370
2003).49673610010750100150200230255
Both52473610010760110140195
of5467361001075080
the3157371001073085130
two33123710010730105155
methods34923710010780125155210260315355
have38723710010755100150195
drawbacks.4092371001075590135210260305350400440465
Variable45823710010765103138168213263293338
selection4945371001074085115160205235265315370
provides5340371001075590140190220275320360
interpretable3157381001073085115160195250285330360405455485530
model37123810010780130185230260
but39973810010750105135
can4157381001074590145
be4327381001075095
extremely4447381001074595125160205285330360410
variable4882381001075095130160205255285330
because5237381001075095140185240280325
it5587381001073060
is3157391001073070
a32523910010745
discrete3322391001075585125170205250280325
process.3672391001075590140185230270310335
Ridge4032391001076090145190235
regression4292391001073580125160205245285315365420
is4737391001073070
a48323910010745
continuous4902391001074595150180210265320370425465
process5392391001075590140185230270310
that3157401001073085130160
shrinks3342401001074095130160215265305
coefficients3672401001074595140170197225270300345400430470
and41674010010745100155
hence43474010010755100155200245
is4617401001073070
more47124010010780130165210
stable.4947401001074070115165195240265
However,52374010010770120195240290335370390
it3157411001073060
does32424110010755105150190
not34574110010755105135
set3617411001074085115
any37574110010745100150
coefficients3932411001074595140170197225270300345400430470
to4427411001073080
045324110010750
and46074110010745100155
hence47874110010755100155200245
does50574110010755105150190
not52724110010755105135
give5432411001074575125170
an31574210010745100
easily3282421001074590130160190240
interpretable3547421001073085115160195250285330360405455485530
model41024210010780130185230260
(Tibshirani,4387421001073090113163203258288323368423453478
1996).48904210010750100151201231256
In5171421001073085
addition52814210010745100155184214244294349
to3157431001073080
the3262431001073085130
methods34174310010780125155210260315355
mentioned37974310010780125180210240290345390445
above,4267431001074595145195240265
in4557431001073085
statistics,4667431001074070115145175215245275320360385
some5077431001074090170215
improvements5317431001073011016520025030034531574410010780125180210250
of3432441001075080
ridge3537441001073565120165210
estimation3772441001074585115145225270300330380435
method42324410010780125155210260315
(Bashtian4572441001073090135175230260290335390
et4987441001074575
al.,50874410010750785166441001072550
2011;52414410010750100150200225
Jurczyk,3157451001073085120165210260310335
2012;35174510010750100150200225
Kibria3767451001076595145180210255
and40474510010745100155
Saleh,42274510010755100130175230255
2012;45074510010750100150200225
Park47574510010755100135186
and49684510010745100155
Yoon,51484510010765108158213238
2011)54114510010750100150200230
have31574610010755100150195
been3377461001075095140195
proposed3597461001075590140195245285330385
in4007461001073085
recent4117461001073580125170225255
years.4397461001075095140175215240
However,46624610010770120195240290335370390
they5077461001073085130180
still5282461001074070100130160
need54674610010755100145200
a31574710010745
large3227471001073075110155200
number34524710010755110190240285320
of3797471001075080
observation39024710010750100140185220270315345375425480
data.44074710010755100130175200
Variable32774810010765103138168213263293338
selection3640481001074085115160205235265315370
is4035481001073070
a41304810010745
non-deterministic42004810010755105160190245290320365400480510565595635665695740
polynomial49654810010755105135185240290370400445475
(NP)31574910010730318749270855610032874911010730
problem3342491101075590140190220265345
whose37124911010775130180220265
search4002491101074085130165210265
space4292491101074095140185230
is4547491101073070
very4642491101075095130180
large4847491101073075110155200
(Amaldi50724911010730100180225255310340
and54374911010745100155
Kann,31585011010765110165220245
1998).34285011010750100150200230255
During37085011010765120155185240285
the4018501101073085130
past41735011010755100140170
decades,43685011010755100145190245290330355
hundreds47485011010755110165220255300355395
of5168501101075080
methods52735011010780125155210260315355
have31585111010755100150195
been3378511101075095140195
proposed3598511101075590140195245285330385
to4008511101073080
solve4113511101074090120170215
this4353511101073085115155
problem,4533511101075590140190220265345370
including4928511101073085130160215270300355400
genetic5353511101074590145190220250295
prediction,3158521101075590135190220265295325375430455
decision36385211010755100145175215245295350
tree4013521101073065110155
prediction,4193521101075590135190220265295325375430455
Bayes46735211010760105155200240
prediction,4938521101075590135190220265295325375430455
least5418521101073075120160190
square3158531101074090145190225270
prediction,3453531101075590135190220265295325375430455
and39335311010745100155
support4113531101074095150205255290320
vector4458531101075095140170220255
machine47385311010780125170225255310355
prediction5118531101075590135190220265295325375430
(Guyon3158541001073095150200250305
and34885410010745100155
Elisseeff,3668541001075585115155195240285315345370
2003),40635410010750100150200230255
and43435410010745100155
all4523541001074575105
these4653541001073085130170215
methods48935410010780125155210260315355
need52735410010755100145200
a54985410010745
huge31585510010755110155200
computation.3383551001074595175230285315360390420470525550
Greedy39585510010765100145190245295
search4278551001074085130165210265
strategy4568551001074070105150180225270320
can4913551001074590145
greatly5083551001074580125170200230280
reduce5388551001073580135190235280
the3158561001073085130
search3313561001074085130165210265
space3603561001074095140185230
and38585610010745100155
thus4038561001073085140180
improve42435610010730110165200250300345
the4613561001073085130
efficiency4768561001074575102130175205250305350400
of5193561001075080
the5298561001073085130
algorithm.5453561001074575120170315857100107356595150230255
The34385710010760115160
greedy3623571001074580125170225275
search3923571001074085130165210265
methods42135710010780125155210260315355
are4593571001074580125
mainly47435710010780125155210240290
divided5058571001075585135165220265320
into5403571001073085115165
three3158581001073085120165210
categories:3393581001074590120165210260295325370410435
forward3853581001072979114189234269324
selection,4202581001074085115160205235265315370395
backward4621581001075095140191265310346400
elimination,5047581001074474104184214269314344374424479504
and55765810010745100155
stepwise3158591001074070115170245275315360
regression3543591001073580125160205245285315365420
analysis.39885910010745100145175225265295335360
Nonetheless,43735910010770120175220250305350380425465505530
there4928591001073085130165210
are5163591001074580125
common5313591001074595175255305360
disadvantages3158601001075585125170225275320375405450495540580
in3763601001073085
these3873601001073085130170215
methods.41136010010780125155210260315355380
Not45186010010770120150
all4693601001074575105
the4823601001073085130
possible49786010010755105145185215265295340
combinations534360100107459517522525531586110010755100130160210265305
of3488611001075080
variables3593611001075095130160205255285330370
are3988611001074580125
taken4138611001073075125170225
into4388611001073085115165
account,4578611001074590135185240295325350
and49536110010745100155
the5133611001073085130
results5288611001073580120175205235275
which31586210010775130160205260
greatly3443621001074580125170200230280
depend37486210010755100155200255310
on40836210010750105
the4213621001073085130
evaluation4368621001074595140170225270300330380435
criterion4828621001074580110140185220250300355
are5208621001074580125
usually5358621001075595150195225255305
not31586310010755105135
the3318631001073085130
optimal34736310010750105135165245290320
solution.3818631001074090120175205235285340365
In4208631001073085
addition,43186310010745100155185215245295350375
the4718631001073085130
variable4873631001075095130160205255285330
transformation,522863100107306511016520523528532031586410010780125155185235290315
such3498641001074095140195
as3718641001074585
principal3828641001075590120175220250305350380
component4233641001074595175230280335380435465
analysis,47236410010745100145175225265295335360
partial51086410010755100135165195240270
least5403641001073075120160190
The4507037010760115160
Institute635703701073085125155185215270300345
of1005703701075080
Remote11107037010760105185235265310
Sensing14457037010755100155195225280325
and17957037010745100155
Digital1975703701076595140170200245275
Earth22757037010755100135165220
(RADI),2520703701073090160225255285310
Chinese4507137010760115145200245285330
Academy8057137010770115160215260340390
of1220713701075080
Sciences,13257137010755100130175230275320360385
No.917357137010770120145195
Dengzhuang19557137010765110165210255310365410465510
South24907137010755105160190245
Road,27607137010760110155210235
Haidian4507237010770115145200230275330
District,805723701076595135165200230275305330
Beijing11607237010760105135160190245290
100094,14757237010750100150200250300325
China18257237010760115145200245
(wljiao@ceode.ac.cn).20957237010730105135160190235285352397442492547592617662707732777832862887
Photogrammetric37306657010755110160190240285320365445525570600635665710
Engineering44656657010755110155185240285330365395450495
&49856657010760
Remote50706657010760105185235265310
Sensing54056657010755100155195225280325
Vol.40466765010765108138163
80,42346765010750100125
No.43846765010771121146
3,4555676501075075
March46556765010785130165210265
2014,49456765010750100150200225
pp.51956765010755110135
261–269.53556765010750100150200250300350375
0099-1112/14/8003–261472968730107501001502002302803303804304655155656006507007518018519019511001
©38277001010767
201439197001010750100150200
American41447001010770150195230260305350405
Society45747001010755105150180225255305
for4904700101073080115
Photogrammetry50457001010755110160190240285320365445525570600635685
and48907109010745100155
Remote50707109010760105185235265310
Sensing54057109010755100155195225280325
doi:45347237010755105135160
10.14358/PERS.80.3.261471972370107501001251752252753253754104655205806356607107607858368619119611011
ASPRS553187016183157240330403
MEMBERS9881870161108191299385468558631
AND1651187016183175267
SUBSCRIBERS,1949187016174165251324410499548634717807880911
CLICK2891187016186165214300390
HERE3312187016196179269352
TO3696187016179171
READ3899187016189173256348
THE4278187016180176259
ENTIRE4569187016183175254303393476
ARTICLE5077187016183173252301387466550