This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
1
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
A hyb id pho og amme y app oach o
a chaeological si es: Block alignmen issues in a
case s udy ( he Roman camp o A Cidadela)
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J.
Abs ac
Pho og amme y is a cos -e ec i e and e sa ile echnique used o he h ee-
dimensional (3D) egis a ion o a chaeological he i age si es. Managing da ase s o
he e ogeneous images in e ms o came a ype, ele a ion pla o m, posi ion o
acquisi ion ime can now be add essed by s uc u e om mo ion (S M) so wa e ia
bundle adjus men in a single block based on collinea i y p inciples. This de elopmen
enables new possibili ies wi h ega d o da a comple eness assu ance o 3D
documen a ion, e en o complex si es wi h occlusi e elemen s and hidden a eas.
Howe e , hyb id pho og amme y in la ge da ase s o en equi es mul iple
pho og amme ic blocks ha mus be p ocessed indi idually and subsequen ly aligned
o ob ain a uni ied poin cloud. In his pape , we discuss he s eps equi ed o
homogenize he in o ma ion and he me hods used o pe o m block alignmen in hese
cases. A case s udy o low-al i ude ae ial pho og amme y wi h se e al came as and
pla o ms is p esen ed o he Roman camp o A Cidadela in NW Spain as a
ep esen a i e example o an a chaeological si e ha is di icul o su ey using a single
pho og amme ic pla o m. The ela i ely la ge expanse o he a ea and he ac ha i is
pa ially co e ed by a p o ec i e s uc u e cons i u e an ideal amewo k o he usion
o mul ipla o m image y. The mos accu a e digi al su ace model (DSM) was ob ained
ia poin -based me hod usion, du ing which subse s a e aligned based on au oma ically
ex ac ed ie poin s (TPs) be ween he dense poin clouds; howe e , poin -based me hod
usion is e y ime consuming. When ha dwa e capabili ies allow, conduc ing he
p ocess in a single block is p e e able, which is a no iceably mo e accu a e p ocedu e
han independen block usion.
Keywo ds
S M, modelling, hidden a eas, ae ial, UAV, pole
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
2
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
1 Resea ch Aims
This pape p esen s an o e all pipeline o hyb id pho og amme y
1
and discusses he
ypical p oblems o in eg a ing mul iple pho og amme ic sou ces; i ocuses p ima ily
on he analysis o he p ocedu es used o pe o m block usion o ob ain a comple e
digi al su ace model (DSM) o an a chaeological si e, bu i also includes
ecommenda ions (e.g., p ac ical ips o image acquisi ion, p ocedu es o homogenize
da a o adiome ic pos -p ocessing) o si ua ions in which mul iple image sou ces mus
be in eg a ed. Finally, an accu acy assessmen is pe o med in he case o he Roman
se lemen o A Cidadela (NW Spain), whose complexi y equi es he use o mul iple
pla o ms (UAV and pole) and se e al came a ypes.
2 In oduc ion
The documen a ion o a chaeological he i age is an essen ial ask in e ms o i s s udy,
dissemina ion, p ese a ion and e en econs uc ion. Recen ly, se e al echniques ha e
been employed o his pu pose, bu he use o lase scanning is conside ed o be one o
he mos p ecise o 3D eco ding (Senol e al., 2017; Be mejo e al., 2018).
Howe e , he g ea es e olu ion in he way cul u al he i age is documen ed has been
mode n pho og amme y, as is e iden in a wide ange o s udies (G een, Be an and
Shapland, 2014; Robleda e al., 2015; Sapi s ein, 2016). Like adi ional s e eoscopic
pho og amme y, s uc u e om mo ion (S M) uses well-de ined geome ical ea u es
1
The e m "hyb id pho og amme y" is used in his con ex o e e o he usion o
mul ipla o m close- ange pho og amme y. In his pape , we e e pa icula ly o low-
al i ude ae ial pho og amme y using a pole and UAV-based pho og amme y.
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
3
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
cap u ed in mul iple images om se e al angula iewpoin s o gene a e a 3D poin
cloud (Wes oby e al., 2012). Howe e , while adi ional pho og amme y p ima ily
de i es he calib a ion pa ame e s o he came a and he ex e io o ien a ion pa ame e s
om well-dis ibu ed g ound con ol poin s (GCPs) and ie poin s (TPs), he S M
app oach simul aneously compu es bo h he ela i e p ojec ion geome y and a se o
spa se 3D poin s (Chiab ando, Donadio and Rinaudo, 2015).
Mode n pho og amme y based on au oma ic image co ela ion (AIC) algo i hms
enables signi ican au oma ion o he p ocess (Remondino e al., 2013); howe e ,
au oma ed so wa e p ocesses also impose high demands on he quali y and geome ic
con igu a ion o images (Tscha e al., 2015). In pa icula , complex objec geome ies
and occlusi e elemen s can pose p oblems o co ela ion algo i hms. Mo eo e ,
because AIC equi es high o e lap and a e y dense image ne wo k o gua an ee
comple eness, he adiome ic a ia ions o objec s in di e en images can gene a e
ma ching ails. In his sense, an imp o emen in he esul s can be expec ed wi h he use
o a uni o m da a sou ce. Howe e , in complex a chaeological pho og amme ic
su eys, geome ic equi emen s canno be ensu ed by using one image sou ce
exclusi ely. Thus, in many cases, only he e ec i e combina ion o mul ipla o m
pho og amme y can gua an ee comple eness.
The S M and AIC algo i hms allow o ien a ion pa ame e s o be compu ed indi idually
o each image, enabling he simul aneous p ocessing o a la ge numbe o images wi h
di e en posi ions o aken by di e en came as (wi h a ied ocal leng hs, image sizes,
e c.) in a single block (Man e dini and Galassi, 2013). The inc easing p ocessing
capabili ies o new IT equipmen educe econs uc ion imes; howe e , la ge ba ches
o images a e becoming inc easingly common due o he use o massi e cap u e sys ems
such as unmanned ae ial ehicles (UAVs), which ha e g ea ly educed he wo k o la ge
su eys (A i and Essa, 2017; Campana, 2017; Field, Wai e and Wandsnide , 2017).
Thus, he ade-o be ween da ase size and p ocessing capaci y emains he main ac o
ha de e mined whe he o wo k wi h he en i e da ase a once o o di ide he images
in o se e al subse s (i.e., indi idual blocks; hese blocks could be de ined by o he
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
4
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
names depending on he p og am used, such as “chunks” o Pho oscan (Agiso , Sain
Pe e sbu g, RU), “p ojec s” o Pix4D (Pix4D SA, Lausanne, CH) and “componen s”
o Reali yCap u e (Cap u ing Reali y, B a isla a, SK)). Alignmen is also essen ial is
when adding an ex a se o images o a block ha is al eady aligned.
Because each indi idual poin cloud is geo e e enced by an independen bundle
adjus men and sepa a e wo k lows, he e is he po en ial o misalignmen s be ween
blocks. In his sense, S M pho og amme y con inues o inhe i om classic
pho og amme y he need o use subsequen me hods o alignmen o ob ain an
accu a ely me ged block. Gene ally, his alignmen p ocess is pe o med using solid
igid ans o ma ions (e.g., simple o a ions, ansla ions o hei combina ion) based on
e e ence elemen s in one o he blocks ha a e used o ob ain he coo dina es o he
homologous poin s in he o he blocks. The me hods mainly di e in e ms o he
e e ence elemen s used o his igid ans o ma ion. Among he mos common op ions
implemen ed in comme cial pho og amme y packages a e he use o manually en e ed
TPs, common images in di e en "chunks" o a dense se o homologous poin s
(au oma ic TPs) ex ac ed by ma ching be ween poin clouds.
Me ging pho og amme ic sou ces is a e y common p ocedu e in many ields (Bemis
e al., 2014; Fugazza e al., 2018), such as hose ela ed o cul u al he i age (Bolognesi
e al., 2014; Seibe on Fock e al., 2017). Howe e , while independen block accu acy
analysis is a equen ly add essed subjec (Noce ino e al., 2013; Ma ínez-Ca icondo
e al., 2018), he accu acy o he alignmen p ocedu es has no i sel been ex ensi ely
s udied, p obably due o a lack o knowledge o he in e nal wo kings o comme cial
so wa e packages, which a e essen ially black boxes.
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
5
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
3 Ma e ials and Me hod
3.1 Desc ip ion o he s udy si e
The camp o A Cidadela (43º 05´ N, 8º 02´ W, al i ude 475 m MAMSL) is loca ed in
Sob ado dos Monxes (A Co uña) in no hwes Spain. This Roman s uc u e si s on a
pla eau and consis s o a mili a y camp om he impe ial e a wi h a ec angula plan
and ounded co ne s. I occupies an a ea o 2.40 ha, al hough he a ea o in e es (AOI)
on which his wo k ocuses is limi ed o a smalle a ea in which he de ensi e sys em
wi h mason y walls a e isible (Fig. 1, zones I and II).
Fig. 1. Roman camp o A Cidadela. a) Loca ion and gene al iew o he camp: b) ou side a ea (I)
and c) a ea oo p o ec i e co e (II).
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
6
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
F om a pho og amme ic su ey pe spec i e, he main ea u e o his se lemen is he
exis ence o an auxilia y s uc u e o p o ec he exca a ed a ea, which is e y common
in a chaeological exca a ions. A i icial ceilings a e used o p o ec s uc u es and walls
ha ha e los hei "o ganic" cha ac e o become agmen s wi h ma e ial condi ions
ha a e di e en om hose hey we e o iginally designed o and used wi h, which
makes hem highly ulne able o en i onmen al condi ions (So ia, Gue e o and Ga cía,
2017). Howe e , hese auxilia y s uc u es ypically pose challenges when modelling
si es because hey do no enable he acquisi ion o ae ial UAV images, and hey make
he acquisi ion o pole images di icul i he heigh o he s uc u e is no su icien .
3.2 Field campaign and pho og amme ic su ey
Con en ional S M pho og amme y equi es high o e lap o ensu e he gene a ion o
he model, while in hyb id pho og amme y (i.e., he combina ion o mul isou ce
image y), he le el o o e lap is e en mo e impo an , especially in he common zone
be ween blocks. The e o e, he planning and image acquisi ion s eps should be
pe o med using geome ic cons ain s o good spa ial in e sec ion and o e lap. The
o e lap o his case s udy (i.e., 70% and 80% ac oss and along he ack, espec i ely)
was de ined in he p io planning phase o ob ain good esul s in he dense ma ching
p ocess.
In Augus 2015, a ligh was conduc ed using a e ical ake-o and landing (VTOL)
o a ing-wing DJI Phan om 3 UAV (www.dji.com). The de ice weighs 1,280 g and has
a speci ic gimbal o s abilize i s in eg a ed came a. The d one’s came a inco po a es a
6.16 × 4.62 mm (1/2.3) CMOS senso ha can cap u e 12.4-Mp images (4,000 × 3,000
pixels). The g ound con ol s a ion and he UAV we e adio-linked and ansmi ed
al i ude, posi ion and s a us da a. To achie e a g ound sample dis ance (GSD) less han
1 cm, he de ined ligh heigh was 26 m.
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
7
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
The digi al single-lens e lex (DSLR) came a used o he a ea unde he oo was a
Canon EOS 550D wi h a 22.3 × 14.9 mm 18-megapixel CMOS senso (5,184 × 3,456
pixels) wi h a pixel size o 4.3 × 4.3 μm. To imp o e he measu emen quali y o he
model, a Canon EF 20-mm /2.8 ixed ocal leng h lens was used. This came a p o ides
highe quali y images han he one in eg a ed in he UAV. To ele a e he came a, i was
necessa y o employ a pole and use a i s -pe son iew sys em (FPV glasses) wi e
connec ed wi h a emo e shoo ing bu on. Table 1 and Fig. 2 summa ize he da a
co esponding o he images o bo h a eas.
Fig. 2. Image cap u e sys ems: a) DJI Phan om 3 UAV, b) came a ele a ion wi h pole. Image
samples o a wall showing he di e ences in GSD: c) d one image and d) DSLR image using he
pole.
Table 1. Main cha ac e is ics o he da ase s used
A ea
Image esolu ion
# o images
Mean heigh
GSD
A ea I
12.4 Mp
422
26.9 m
0.94 cm/pix
A ea II
18 Mp
2,769
3.2 m
0.06 cm/pix
A To al S a ion was used o ob ain GCPs and checkpoin s in he ou e and indoo a eas.
An o e lap zone common o he a ea unde he oo and ae ial images was es ablished
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
8
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
(app ox. 3 m in wid h) along he edge o he oo o ensu e image coupling. Figu e 3a
shows he loca ion o he manually en e ed TPs ha we e used o es he in luence o
hei use on he accu acy o he DSM.
Fig. 3. a) Dis ibu ion o GCPs, checkpoin s and manually en e ed TPs on he si e; b) To al S a ion
R-326 used o collec he poin s; c) GCPs iden i ied wi h a ge s easily isible om he ai ; and d)
checkpoin s and TPs di ec ly ma ked o e clea ly iden i iable elemen s in he dense poin cloud.
3.3 Gene al wo k low o hyb id pho og amme y
The gene al wo k low o hyb id pho og amme y p ocessing (Fig. 4) s a s wi h he
a o emen ioned planning phase o image acquisi ion. The planning mus no only
gua an ee he desi ed GSD and a good o e lap be ween he images o each block bu
also conside he zone o o e lap be ween blocks. The 3D modelling p ocess i sel ,
using so wa e based on S M algo i hms (Remondino e al., 2017), in ol es he
ollowing s eps: (1) au oma ic co ela ion, (2) hie a chical o ien a ion, and (3) au oma ic
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
9
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
3D dense su ace ex ac ion. In s age (1), he so wa e sea ches o common poin s on
he pho og aphs and ma ches hem; i also inds he posi ion o he came a o each
image and e ines he came a calib a ion pa ame e s (2). As a esul , a se o came a
posi ions and a spa se poin cloud a e o med. Based on he es ima ed came a posi ions,
he p og am calcula es he dep h in o ma ion o each came a, which is combined in o a
single dense poin cloud (3). The las s eps o he p ocess in ol e he gene a ion o he
mesh om he poin cloud and he c ea ion o he ex u e iles, which p oduce he inal
DSM when combined.
Fig. 4. Gene al wo k low o hyb id pho og amme y p ocessing
Al hough S M-based pho og amme y allows some lexibili y in he geome y o he
images, he geome ies canno be e y di e en i o ien a ion p ocessing is pe o med
simul aneously in a single block. In o he cases, wide baselines can cause image
alignmen ailu es (Hidaya and Cahyono, 2016). To a oid his issue, common poin s
ha link bo h ypes o image blocks should be in oduced. The in oduc ion o GCPs
be o e he hie a chical o ien a ion and dense econs uc ion p ocesses is especially
impo an when wo king wi h se e al di e en senso s and pla o ms.
Any S M so wa e equi es a minimum o 3 GCPs o pe o m pho og amme ic block
adjus men , as well as scaling, o a ing and loca ing he model. Each one o hose poin s
should be iden i ied in a leas wo images. Al hough se e al s udies ha e discussed his
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
16
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
4.2 Gene al wo k low o hyb id pho og amme y
The DSMs o each a ea (Fig. 5) we e gene a ed ollowing he desc ibed wo k low. The
es ima ed posi ions o he GCPs in he DSMs gene a ed o a eas I and II clea ly di e
in e ms o he oo mean squa e e o (RMSE) o he adjus men o he models (Table
3), mainly due o he di e ences in he achie able g ound esolu ions. In he ae ial
images (I), a dense poin cloud wi h 9.330 mm/pix was gene a ed, which con as s he
0.602 mm/pix in he a ea unde oo (II). As his me hod o quali y checking nei he
equi es no uses independen measu emen s, he alue should be analysed in e ms o
in e nal p ecision a he han o e all accu acy.
Fig. 5. Image posi ions (black do s) and numbe o o e lapped images in bo h a eas o he
a chaeological si e: a) ou side a ea (I) and b) a ea unde p o ec i e co e (II). Dense poin clouds
gene a ed o : c) he ou side a ea (I) and d) he a ea unde p o ec i e co e (II). The unmodelled
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
17
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
cen al a ea co esponds o a la unexca a ed a ea, which was no included because i has no
a chaeological in e es . Comple e 3D a ailable in Ske ch ab: h ps://sk b.ly/6zG7 .
Table 3. RMSE alues based on GCP adjus men .
RMSE X
(cm)
RMSE Y
(cm)
RMSE Z
(cm)
To al RMSE
(cm)
DSM - A ea I (UAV)
1.5572
2.5004
3.2200
4.3641
DSM - A ea II (pole)
11.3050
6.0361
13.3609
18.5135
Single Block DSM – En i e A ea
6.2587
8.3401
19.2046
21.8528
4.3 Block alignmen me hods: Accu acy assessmen
When p ocessing a single block, he S M so wa e uses wha e e cons ain s i has
(coo dina es o GCPs, TPs, ix dis ances and p e-calib a ed came a pa ame e s) o
op imize he spa se poin cloud, came a spa ial posi ions, and o ien a ions and imp o e
came a calib a ion. By emo ing ce ain cons ain s (such as poo TPs), he alignmen
esul is op imized (a new "bes - i " model is calcula ed by bundle adjus men ), and he
o e all accu acy is inc eased. In he case o a single block, all o hese cons ain s a e
linked because hey ha e been compa ed h ough a simul aneous adjus men and he
TPs (gene ally he mos impo an cons ain in S M pho og amme y) o m a single
in e linked whole. The bundle adjus men based on he collinea i y equa ion allows o
di e en came as o be o ien ed in a single p ocess because he calib a ion pa ame e s
can be calcula ed o subg oups o images in he same block o o each image
indi idually. The esul s (Fig. 6) show ha he o e all accu acy o he DSM can be
signi ican ly highe using he same GCPs in single-block p ocessing (3.77-cm RMSE)
because i a oids p opaga ing block alignmen e o s. The o al ime does no
signi ican ly di e om ha equi ed o p ocess blocks sepa a ely and o pe o m he
alignmen and usion. Thus, conduc ing single-block p ocessing is ecommended bu
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
18
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
limi ed by he DSM size and he capaci y o he compu e o compu e he ba ch o
images simul aneously.
Fig. 6. Accu acy assessmen wi h he dis ances be ween checkpoin s in bo h blocks. The g aph
compa es he RMSE alues o each ial.
When a poin -based alignmen ope a ion is used, S M so wa e s a s he image-
ma ching p ocess o images om all blocks being aligned. This me hod is he mos
ime-in ensi e alignmen me hod because i equi es he analysis o he wo dense poin
clouds in sea ch o common poin s. The TPs ound a e exclusi ely used in he block
alignmen and a e disca ded a e he ope a ion.
In he case o ma ke -based alignmen , when a block is being aligned wi h ano he
e e ence block, he es ima ed ma ke coo dina es o he e e ence block a e used. The
union o mul iple consecu i e blocks exclusi ely using ma ke s can lead o addi ional
accumula ed e o s. The e o e, his alignmen me hod is ecommended only when
adequa e ma ke accu acy can be gua an eed. The numbe o ma ke s mus also be
su icien o gua an ee ha he p og am can op imize he new posi ion o he block
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
19
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
using leas squa es. Based on he dis ance accu acy assessmen , adding 8 o 12 well-
dis ibu ed, manually en e ed TPs in he o e lap a ea can no ably imp o e he accu acy
o he DSM.
The alignmen o blocks using he came a-based me hod equi es wo king wi h blocks
ha sha e common pho og aphs. The accu acy assessmen shows simila esul s wi h
ega d o alignmen using came as and poin s (5.09-cm RMSE and 4.60-cm RMSE,
espec i ely). Howe e , he came a me hod akes a ew seconds a he han he up o 6
hou s equi ed o he poin -based me hod. Including common pho og aphs a he
bounda y be ween blocks in each is essen ial o alignmen using he came a me hod
and is also impo an o TP gene a ion when using poin -based alignmen .
5 Conclusions
In eg a ed 3D modelling using di e en pla o ms is expec ed o con ibu e o digi al
documen a ion echnology o he spa ial analysis o complex a chaeological si es.
Howe e , good planning o he a eas o o e lap when aking pho og aphs is essen ial o
ensu e he comple eness and accu acy o he DSM.
The a ying g ound esolu ion o images can be pa ially add essed using simpli ica ion
and il e ing algo i hms o homogenize he da a. Fo ha ma e , p og ams used in
pho og amme y o en p o ide solu ions o homogenizing ex u es o achie e mo e
uni o m esul s.
When wo king wi h la ge da ase s ha exceed he p ocessing capabili ies o he
equipmen , o when i is necessa y o add an ex a se o images o an al eady p ocessed
p ojec , mul iple blocks may be equi ed. In hose cases, co ec block alignmen is
c i ical when a emp ing o cons uc a single poin cloud. The poin -based alignmen
me hod yields be e esul s in e ms o accu acy because i uses a highe numbe o TPs
de i ed om a new ma ching p ocess be ween he poin clouds. In con as , ma ke -
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
20
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
based and came a-based me hods pe o m block alignmen mo e quickly. Came a-based
alignmen equi es some common pho og aphs be ween adjacen blocks; clea ly
iden i iable poin s should be used when using en e ed ma ke s as TPs because i
equi es high p ecision o add poin s in bo h blocks. Based on he esul s, he numbe o
TPs should be app oxima ely 8. Adding mo e TPs does no no ably imp o e he
accu acy, and i can be a ime-consuming ask o iden i y a highe numbe o poin s o e
he common a ea o he blocks o be me ged. No ably, d awing conclusions abou he
alignmen wi h ma ke s based on he abo e es s may no be app op ia e because his
p ac ical case is oo simple wi h ega d o he numbe o "chunks" (only wo blocks).
Fu he s udies should ocus on he analysis o he possible accumula ion o e o s when
mul iple consecu i e blocks a e aligned.
Acknowledgemen s
The s udy was suppo ed by Xun a de Galicia unde he "Financial aid o he
consolida ion and s uc u e o compe i i e uni s o in es iga ion in he uni e si ies o
he Uni e si y Galician Sys em (2016-18)" g an Re . ED431B 2016/030 and Re .
ED341D R2016/023. The au ho s hank all p ojec pa ne s and collabo a o s. We
would especially like o hank D . San iago Ma ínez, PhD, o he echnical suppo and
IRIS UAV Se ices S.L. o conduc ing he ligh .
6 Re e ences
A i , R. and Essa, K. (2017) ‘E ol ing echniques o documen a ion o a wo ld he i age
si e in Laho e’, in The In e na ional A chi es o he Pho og amme y, Remo e Sensing
and Spa ial In o ma ion Sciences, 26 h In e na ional CIPA Symposium 2017. O awa,
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
21
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
Canada, pp. 33–40. doi: 10.5194/isp s-a chi es-XLII-2-W5-33-2017.
Bemis, S. e al. (2014) ‘G ound-based and UAV-Based pho og amme y : A mul i-scale
, high- esolu ion mapping ool o s uc u al geology and paleoseismology’, Jou nal o
S uc u al Geology. Else ie L d, 69, pp. 163–178. doi: 10.1016/j.jsg.2014.10.007.
Be mejo, L. e al. (2018) ‘Assessing he accu acy o 3D GPR esul s by compa ing
hem o 3D lase scanne models : he case s udy o he a chaeological si e o Peluda
Ca e ( Sie a de A apue ca , Bu gos , Spain )’, in Geophysical Resea ch Abs ac s.
EGU Gene al Assembly 2018.
Bolognesi, M. e al. (2014) ‘Accu acy o Cul u al He i age 3D Models by RPAS and
Te es ial Pho og amme y’, in ISPRS Technical Commission V Symposium. Ri a del
Ga da, I aly ACCURACY: The In e na ional A chi es o he Pho og amme y, Remo e
Sensing and Spa ial In o ma ion Sciences, pp. 23–25. doi: 10.5194/isp sa chi es-XL-5-
113-2014.
Campana, S. (2017) ‘D ones in A chaeology. S a e-o - he-a and Fu u e Pe spec i es’,
A chaeological P ospec ion, 24(4), pp. 275–296. doi: 10.1002/a p.1569.
Chiab ando, F., Donadio, E. and Rinaudo, F. (2015) ‘S M o o hopho o gene a ion:
Awinning app oach o cul u al he i age knowledge’, In e na ional A chi es o he
Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sciences - ISPRS A chi es,
40(5W7), pp. 91–98. doi: 10.5194/isp sa chi es-XL-5-W7-91-2015.
Field, S., Wai e, M. and Wandsnide , L. (2017) ‘The u ili y o UAVs o a chaeological
su ace su ey: A compa a i e s udy’, Jou nal o A chaeological Science: Repo s.
Else ie , 13(Ap il), pp. 577–582. doi: 10.1016/j.jas ep.2017.05.006.
Fugazza, D. e al. (2018) ‘Combina ion o UAV and e es ial pho og amme y o
assess apid glacie e olu ion and map glacie haza ds’, Na u al Haza ds and Ea h
Sys em Sciences, 18, pp. 1055–1071.
Golds ein, E. e al. (2015) ‘G ound con ol poin equi emen s o s uc u e- om-
mo ion de i ed opog aphy in low- slope coas al en i onmen s’, Pee J P eP in s. doi:
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
22
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
10.7287/pee j.p ep in s.1444 1 |.
G een, S., Be an, A. and Shapland, M. (2014) ‘A compa a i e assessmen o s uc u e
om mo ion me hods o a chaeological esea ch’, Jou nal o A chaeological Science.
Else ie L d, 46(1), pp. 173–181. doi: 10.1016/j.jas.2014.02.030.
Hidaya , H. and Cahyono, A. (2016) ‘Combined ae ial and e es ial images o
comple e 3D documen a ion o Singosa i Temple based on S uc u e om Mo ion
algo i hm’, in IOP Con e ence Se ies: Ea h and En i onmen al Science. IOP
Publishing, pp. 0–11. doi: 10.1088/1755-1315/47/1/012004.
Lee, D. T. and Schach e , B. J. (1980) ‘Two algo i hms o cons uc ing a Delaunay
iangula ion’, In e na ional Jou nal o Compu e & In o ma ion Sciences, 9(3), pp.
219–242. doi: 10.1007/BF00977785.
Man e dini, A. M. and Galassi, M. (2013) ‘Assessmen o 3D Recons uc ions o
Cul u al He i age using Digi al Technologies’, ISPRS - In e na ional A chi es o he
Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sciences, XL-5/W1(Feb),
pp. 167–174. doi: 10.5194/isp sa chi es-XL-5-W1-167-2013.
Ma ínez-Ca icondo, P. e al. (2018) ‘Assessmen o UAV-pho og amme ic mapping
accu acy based on a ia ion o g ound con ol poin s’, In e na ional Jou nal o Applied
Ea h Obse a ion and Geoin o ma ion. Else ie , 72(May), pp. 1–10. doi:
10.1016/j.jag.2018.05.015.
Noce ino, E. e al. (2013) ‘Accu acy and block de o ma ion analysis in au oma ic UAV
and e es ial pho og amme y - lesson lea n ’, in ISPRS Annals o he
Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sciences, XXIV In e na ional
CIPA Symposium. S asbou g, F ance, pp. 2–6. doi: 10.5194/isp sannals-II-5-W1-203-
2013.
Pales ini, C. and Basso, A. (2017) ‘Geoma ics as a su ey ool o documen and enhance
he cul u al and landscaped he i age o he monumen al complexes in he moun ains o
Ab uzzo’, in The In e na ional A chi es o he Pho og amme y, Remo e Sensing and
Spa ial In o ma ion Sciences, 2017 GEOMATICS & RESTORATION – Conse a ion o
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
23
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
Cul u al He i age in he Digi al E a. Flo ence, I aly, pp. 373–380. doi: 10.5194/isp s-
A chi es-XLII-5-W1-373-2017.
Remondino, F. e al. (2013) ‘Dense image ma ching: Compa isons and analyses’, in
P oceedings o he Digi alHe i age 2013 - Fede a ing he 19 h In ’l VSMM, 10 h
Eu og aphics GCH, and 2nd UNESCO Memo y o he Wo ld Con e ences, Plus Special
Sessions omCAA, A queologica 2.0 e al., pp. 47–54. doi:
10.1109/Digi alHe i age.2013.6743712.
Remondino, F. e al. (2017) ‘A c i ical e iew o au oma ed pho og amme ic
p ocessing o la ge da ase s’, in The In e na ional A chi es o he Pho og amme y,
Remo e Sensing and Spa ial In o ma ion Sciences, 26 h In e na ional CIPA Symposium
2017. O awa, Canada, pp. 591–599. doi: 10.5194/isp s-a chi es-XLII-2-W5-591-2017.
Robleda, P. G. e al. (2015) ‘Modeling and accu acy assessmen o 3D- i ual
econs uc ion in cul u al he i age using low-cos pho og amme y: Su eying o he
“san a ma ía azogue” chu ch’s on ’, In e na ional A chi es o he Pho og amme y,
Remo e Sensing and Spa ial In o ma ion Sciences - ISPRS A chi es. Else ie B.V.,
40(5W4), pp. 263–270. doi: 10.5194/isp sa chi es-XL-5-W4-263-2015.
Sapi s ein, P. (2016) ‘Accu a e measu emen wi h pho og amme y a la ge si es’,
Jou nal o A chaeological Science. Else ie L d, 66, pp. 137–145. doi:
10.1016/j.jas.2016.01.002.
Seibe on Fock, S. e al. (2017) ‘Pipeline o Recons uc ion and Visualiza ion o
Unde wa e A chaeology Si es using Pho og amme y Recons uc ion Image
P ocessing’, in 32nd In e na ional Con e ence on Compu e s and Thei Applica ions.
Honolulu.
Senol, H. I. e al. (2017) ‘3D Modeling o a Bazaa in Ancien Ha an Ci y Using Lase
Scanning Technique’, in 4 h In e na ional GeoAd ances Wo kshop. Ka abuk, Tu key:
The In e na ional A chi es o he Pho og amme y, Remo e Sensing and Spa ial
In o ma ion Sciences, pp. 99–101. doi: 10.5194/isp s-a chi es-XLII-4-W6-99-2017.
So ia, F. J., Gue e o, L. F. and Ga cía, A. B. (2017) ‘P o ec i e Roo Sys ems o
This is an accep ed e sion o a pape published in he Jou nal o Cul u al He i age
________________________________
24
________________________________
EDITOR FULL VERSION AVAILABLE ON:
h ps://www.sciencedi ec .com/science/a icle/pii/S1296207418305089
A za-Ga cía, M., Gil-Docampo, M., & O iz-Sanz, J. (2019). A hyb id pho og amme y app oach o a chaeological
si es: Block alignmen issues in a case s udy ( he Roman camp o A Cidadela). Jou nal o Cul u al He i age. DOI:
h ps://doi.o g/10.1016/j.culhe .2019.01.001
A cheological Si es in Mexico’, WIT T ansac ions on The Buil En i onmen , 171, pp.
225–236. doi: 10.2495/STR170201.
Tonkin, T. and Midgley, N. (2016) ‘G ound-Con ol Ne wo ks o Image Based Su ace
Recons uc ion : An In es iga ion o Op imum Su ey Designs Using UAV De i ed
Image y and S uc u e- om-Mo ion Pho og amme y’, Remo e Sensing, 8(786), pp. 16–
19. doi: 10.3390/ s8090786.
To es-Ma ínez, J. A. e al. (2015) ‘A Mul i-da a sou ce and mul i-senso app oach o
he 3d econs uc ion and isualiza ion o a complex a chaelogical si e: The case s udy
o olmo de mina eda’, In e na ional A chi es o he Pho og amme y, Remo e Sensing
and Spa ial In o ma ion Sciences - ISPRS A chi es, 40(5W4), pp. 37–44. doi:
10.5194/isp sa chi es-XL-5-W4-37-2015.
Tscha , A. e al. (2015) ‘On he Use o Ua s in Mining and A chaeology - Geo-
Accu a e 3D Recons uc ions Using Va ious Pla o ms and Te es ial Views’, ISPRS
Annals o Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sciences, II-1/W1,
pp. 15–22. doi: 10.5194/isp sannals-II-1-W1-15-2015.
Wes oby, M. J. e al. (2012) ‘“S uc u e- om-Mo ion” pho og amme y: A low-cos ,
e ec i e ool o geoscience applica ions’, Geomo phology. Else ie B.V., 179, pp.
300–314. doi: 10.1016/j.geomo ph.2012.08.021.
