COMPARISON OF 3D REALITY CAPTURE TECHNOLOGIES FOR THE SURVEY OF STONE WALLS

P oceedings o he 8 h In e na ional Cong ess on A chaeology, Compu e G aphics, Cul u al He i age and Inno a ion ‘ARQUEOLÓGICA 2.0’ in Valencia (Spain), Sep . 5 – 7, 2016 DOI: h p://dx.doi.o g/10.4995/a queologica8.2016.2582 Recei ed: 26/01/2016 Accep ed: 18/05/2016 * Co esponding Au ho : En ique Vale o, e. ale [email protected] This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0) EDITORIAL UNIVERSITAT POLITÈCNICA DE VALÈNCIA COMPARISON OF 3D REALITY CAPTURE TECHNOLOGIES FOR THE SURVEY OF STONE WALLS COMPARATIVA DE TECNOLOGÍAS 3D PARA LA INSPECCIÓN DE MUROS DE PIEDRA En ique Vale oa,*, Alan Fo s e a, F édé ic Boschéa, Lyn Wilsonb, Alick Leslieb a Royal Academy o Enginee ing Cen e o Excellence in Sus ainable Building Design, He io -Wa Uni e si y, EH14 4AS, Edinbu gh, UK. [email p o ec ed]; a.m. o s[email p o ec ed]; [email protected] b Sou h Gyle Conse a ion Cen e, His o ic En i onmen Sco land, EH12 9EB, Edinbu gh, UK. [email p o ec ed]; [email p o ec ed] Abs ac : The main enance o he ex e nal ab ic o his o ic buildings cons i u es a la ge po ion o o e all building li e cycle cos s. Ad anced eali y cap u e and da a p ocessing echnologies ha e he po en ial o ans o m exis ing su ey p ac ice, p o iding su eyo s wi h objec i e da a pe aining o building ab ic, in a mo e apid ( equen ), sa e and cos -e ec i e manne . In his pape , we p esen a unique e alua ion o se e al Te es ial Lase Scanning (TLS) and pho og amme ic (PG) sys ems ha assess hei ela i e s eng hs o he su ey o s one walls. The assessmen is conduc ed using an his o ic building selec ed o i s ep esen a i eness o o m, ab ic and condi ion. The wo k conside s pe o mance in e ms o da a accu acy and p ecision, da a comple eness, and p ocess e iciency. The esul s show ha , while TLS p o ides good geome ic da a o gene a e accu a e and aluable 3D models, he quali y o PG econs uc ions can be also be su icien in such con ex s. And conside ing he ela i ely low-cos and po abili y o mode n digi al came as compa ed o lase scanne s, pho og amme y can cons i u e a ealis ic al e na i e o TLS. In addi ion, moun ing a came a on a UAV could u he sol e access issues, p e en ing he need o any addi ional in as uc u e (e.g. sca olding), which would be equi ed when employing TLS. Howe e , a lesson d awn om his wo k is ha e ec i e acquisi ion o pho og amme ic da a equi es ca e ul planning o selec he app op ia e came a se ings and pic u e densi y (and loca ions) o ensu e accu a e and eliable pho og amme ic econs uc ion. This p ocess may be e e ed o as: Planning o Pho og amme y (P4P). Key wo ds: i ual a chaeology, his o ic buildings, cul u al he i age, documen a ion, su ey, 3D econs uc ion, lase scanning, pho og amme y Resumen: Las labo es de man enimien o de la es uc u a ex e na de edi icios his ó icos cons i uyen una g an pa e de los cos es asociados al ciclo de ida de es os. Las ecnologías empleadas pa a la adquisición de da os es uc u ales y su p ocesamien o ienen el po encial de ans o ma , en g an medida, las a eas de inspección adicionales, p opo cionando in o mación ela i a al edi icio de una mane a más ápida, segu a y económica. En es e abajo, se e alúan escáne es láse y sis emas basados en o og ame ía, demos ando su solidez pa a la inspección de pa edes de pied a. Es a e aluación se ha lle ado a cabo en un edi icio his ó ico ep esen a i o y en un buen es ado de conse ación. En el a ículo, se compa a la ac uación de los di e en es disposi i os en é minos de p ecisión, epe ibilidad, densidad de da os y e iciencia del p oceso. Los esul ados mues an que, mien as que los escáne es p opo cionan buena in o mación geomé ica y pe mi en la gene ación de modelos idimensionales p ecisos, la calidad de las econs ucciones a pa i de o og ame ía ambién pueden se su icien es en de e minados con ex os. Conside ando, además, el ela i o bajo cos e y la po abilidad de las cáma as digi ales mode nas en compa ación con los escáne es láse , la o og ame ía puede cons i ui una al e na i a eal a es os úl imos. Además, la ins alación de cáma as en ehículos aé eos no ipulados (d ones) puede esol e p oblemas de accesibilidad y minimiza el uso de in aes uc u as adicionales (como andamiajes), las cuales son empleadas en la digi alización median e escáne es. Sin emba go, una lección que se desp ende de es os expe imen os es que pa a consegui una e ec i a adquisición de da os o og amé icos se equie e una cuidada plani icación y una selección ap opiada de los ajus es de la cáma a, así como su co ec o posicionamien o en la escena, con el obje i o de consegui una adecuada densidad de imágenes y ob ene una econs ucción 3D p ecisa. Es e p oceso se denomina Plani icación pa a la Fo og ame ía. 14 Vale o, Fo s e , Bosche, Wilson, Leslie, 2016. This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0) EDITORIAL UNIVERSITAT POLITÈCNICA DE VALÈNCIA Palab as cla e: a queología i ual, edi icios his ó icos, pa imonio, documen ación, inspección, econs ucción 3D, escáne es láse , o og ame ía 1. In oduc ion The main enance o he ex e nal ab ic o his o ic buildings cons i u es a la ge po ion o o e all building li e cycle cos s. In Sco land alone, i is es ima ed ha he e a e app oxima ely 400,000 adi ional (p e 1919) buildings ep esen ing 20% o building s ock and cons i u ing an impo an po ion o cons uc ion spendings (Muelle e al. 2014). “Repai , main enance and imp o emen accoun s o app oxima ely 46% o he o al cons uc ion indus y ou pu , o which one hi d ela es di ec ly o p e 1919 buildings (accoun ing o app oxima ely £1.2 billion)” (His o ic Sco land 2011). En i onmen al clima ic condi ions and associa ed wea he ing ac as ‘agencies o ma e ials change’, ha lead o he de e io a ion o building ab ic. A es ing his de e io a ion equi es egula , app op ia e main enance. This ope a ion is becoming mo e one ous due o clima e change and mo e speci ically inc eased incidence o ex eme ain all in Sco land. P ac ically, highe le els o mois u e can be co ela ed wi h accele a ed de e io a ion in po ous building ma e ials. P e alen decay mechanism in mason y include inc eased se e i y and equency in eeze- haw cycling and spalling e en s, and biological g ow h associa ed wi h long e m sa u a ion o ab ic. Logically, inc eased agg essi i y o en i onmen al condi ions equi es highe equency e alua ion o condi ion o he ex e nal e elope o buildings. Main enance decisions a e made ollowing de ailed su eys o ab ic. Ad anced eali y cap u e and da a p ocessing echnologies ha e he po en ial o ans o m exis ing su ey p ac ice, p o iding su eyo s wi h objec i e da a pe aining o building ab ic, in a mo e apid ( equen ), sa e and cos -e ec i e manne . In pa icula , g ea e olume and be e accu acy o in o ma ion can suppo su eyo s in ensu ing objec i i y and consis ency when assessing condi ion. This is pa amoun as hese su eys a e u ilised o de elop epai s a egies. Su eying building ab ic encompasses many dispa a e and in e connec ed componen s ha equi e he acquisi ion o a ious ypes o da a. P incipally, amongs in o ma ion is cons uc ion o m and ma e ials. Bo h o hese i s o de su ey objec i es ely upon speci ic da a including he dimensions/geome y and ex u e o he ab ic componen s. Ad anced h ee-dimensional and isual eali y cap u e echnologies ha e been de eloped in he las ew decades o acqui e such da a. These echnologies mainly include: pho og amme y and lase scanning. Bo h can deli e dense 3D poin clouds o scenes, possibly augmen ed wi h colou in o ma ion. Pho og amme ic sys ems a e based on he use o came as and he p inciple o s e eo ision, while lase scanning is a mo e ecen echnology ha is based on he con olled command o lase beams and ime-o - ligh p inciples o de e mine he dis ance o an objec . Despi e he quali y o he da a deli e ed by lase scanne s, hei use is limi ed by he need o se up he scanne on he g ound o a s able pla o m a mul iple loca ions ha p oduce a comple e 3D model when indi idual scans a e egis e ed oge he . Pho og amme ic sys ems ha e shown g ea po en ial o esol e access- ela ed issues bu hei pe o mance can be signi ican ly limi ed by he ex u e and/o ype o ma e ial being scanned. Ano he limi a ion o pho og amme ic sys ems is ha he 3D da a hey p o ide has unknown scale; con ol ne wo ks o known dimensions in he da a a e necessa y o con e he da a o me ic scale, which ine i ably in oduces addi ional e o . The di e en s eng hs and weaknesses o bo h sys ems ha e led esea che s o de e mine which solu ion is mos app op ia e in a ious con ex s. In his pape , we p esen a unique e alua ion o se e al Te es ial Lase Scanning (TLS) and pho og amme ic (PG) sys ems ha a e compa ed o assess hei ela i e s eng hs o he su ey o s one walls. The assessmen is conduc ed using an his o ic building selec ed o i s ep esen a i eness o o m, ab ic and condi ion. The wo k conside s pe o mance in e ms o da a accu acy and p ecision, da a comple eness, and p ocess e iciency. This pape is s uc u ed as ollows: Sec ion 2 con ains a e iew o p io wo ks on he use o lase scanning and pho og amme y o he su ey o his o ic buildings, leading o he iden i ica ion o he knowledge gap ha his esea ch con ibu es o ill. Sec ion 3 desc ibes he echnologies assessed and he me hodology ollowed o compa e hem. Sec ion 4 p esen s he on-si e da a acquisi ion and econs uc ion s eps. Sec ion 5 epo s and analyses he esul s on he compa ison o he da a ob ained o he di e en sys ems. Finally, Sec ion 6 p esen s he conclusions and u u e wo ks. 2. Backg ound Lase scanning is a ela i ely ecen echnology ha cons i u es a e olu ion in land and building su eying. Wi hin he con ex o his o ic monumen su ey, no ewo hy examples o he use o TLS include he wo k o Wilson e al. (2013) who illus a e he ad an ages o TLS o he su ey o la ge and complex his o ic monumen s ia case s udies o UNESCO Wo ld He i age Si es. Ca daci e al. (2011) show ha TLS p o ides signi ican alue compa ed o adi ional manual su ey. They also show how CAD models gene a ed om he da a can be success ully used o s uc u al analysis using he Fini e Elemen Me hod (FEM). Ne ley e al. (2013) use TLS and ligh De ec ion and Ranging (LiDAR) – i.e. ae ial lase scanning – o ob ain a pho o ealis ic geospa ial model o he his o ic quayside a Co ehele Quay in eg a ed in an accu a e Digi al Ele a ion Map (DEM) in o de o assess he po en ial impac o ising sea le els esul ing om clima e change. Temize e al. (2013) show he alue o TLS o su ey unde g ound s uc u es like he Byzan ine cis e n si ua ed benea h he cou o he Sa nicli Han building. They also in es iga e he impac o a ious le els o poin il e ing on he accu acy o he mesh p oduced om he da a. 15 COMPARISON OF 3D REALITY CAPTURE TECHNOLOGIES FOR THE SURVEY OF STONE WALLS This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0) EDITORIAL UNIVERSITAT POLITÈCNICA DE VALÈNCIA Pho og amme y is ano he well-es ablished me hod o ob aining 3D eco ds o his o ic monumen s as he e iew o Ogleby (1995) o example shows. I s use ou da es ha o TLS by decades, almos a cen u y (Sche me ho n 1940). Howe e , signi ican p og ess has been made in he las wo decades, bo h in e ms o ha dwa e and so wa e, ha has enabled signi ican au oma ion o PG da a acquisi ion and p ocessing o 3D model econs uc ion. The de elopmen o obus au oma ed ea u e de ec ion and ma ching in digi al images, (e.g. SIFT (Lowe 1999) o SURF (Bay e al. 2006) ea u es), as well as dense ma ching app oaches (S echa e al. 2003) ha e d ama ically imp o ed, e en au oma ed he image p ocessing s age. Fu he mo e, high- esolu ion and po able digi al came as a e now widely a ailable a a ela i ely low cos , meaning ha pho og amme y can be easily deployed o p oduce dense ex u ed 3D poin clouds and meshes. Thanks o he ligh weigh o mode n came as, single-came a pho og amme y has also shown g ea po en ial o esol e access- ela ed cons ain s. Rega ding he ield o his o ic monumen su ey, ema kable examples o he use o pho og amme y include he wo k o Cappellini e al. (CAPELLINI, V., STEFANI, C., NONY, N. and DE LUCA, L., 2012) who apply i o p oduce 3D models o monumen s ha a e used o gene a e 2.5D o hopho os o walls. Using he example o Roman walls, hese o hopho os a e employed o conduc he seman ic anno a ion o he opus o di e en sec ions o he wall. Le ma and Mui (2014) p esen a compa ison o lase scanning wi h pho og amme y, econs uc ing a 3D model by means o Visual S uc u e o Mo ion (VS M) and Pa ch o Clus e based Mul i View S e eo So wa e (PMVS/CMVS), and concluded ha o accu a e documen a ion o ca ed de ailing on s one, pho og amme ic echniques p o ided he g ea es le el o lexibili y and eliabili y, al hough he wo echniques used in combina ion may p o ide he bes esul s. Howe e , PG sys ems also ha e limi a ions. Fo example, hey a e no obus o a ying ligh ing condi ions and ex u e-poo o e lec i e ma e ials (Tho nbush and Viles 2008). Fu he mo e, hei p ecision quickly d ops wi h he dis ance o he came a o he a ge scene when compa ed o TLS (Rasz o i s and Do ninge 2013). Also, single-came a PG sys ems p o ide 3D econs uc ions only up o scale, and equi e he use o use su ey ne wo ks o ex ac known dimensions in he images o adequa ely scale he econs uc ions (Ska la os and Kipa issi 2012). Wi h ega d o access p o ision, PG and TLS sys ems can bo h be moun ed on poles o ex endable ipods ha enable posi ioning hem a heigh s up o app oxima ely 5m. Fo g ea e heigh s, ae ial solu ions can hen be conside ed. In ac , ae ial pho og amme y and LiDAR (i.e. ae ial lase scanning) sys ems ha e long been conduc ed using planes o helicop e s wi h he senso s moun ed unde hem. Howe e , hese solu ions do no allow he acquisi ion o images a close dis ances, which limi s he accu acy and densi y o he esul ing 3D econs uc ions. Recen and apid de elopmen s o Unmanned Ae ial Vehicles (UAVs), in pa icula helicop e - ype UAVs, o e new pla o ms ha a e pa icula ly sui ed o ligh -weigh PG sys ems. PG came as moun ed on helicop e UAV enable he acquisi ion o images a close anges a any heigh , p o iding a se ious ad an age o e lase scanning sys ems in e ms o access p o ision. I should be no ed ha he e a e a emp s o moun lase scanning sys ems on helicop e UAVs bu he weigh o he scanne s equi es e y la ge UAVs and ensu ing ha he acqui ed da a has high p ecision and accu acy emains a g ea challenge. UAVs ha e al eady been conside ed o use in he applica ion o pho og amme y in he con ex o his o ic monumen s. Fo example, Püschel e al. (2008) p opose he use o e es ial and UAV pic u es o c ea e an accu a e 3D model o Cas le Landenbe g. Remondino e al. (2011) e iew he di e en s ages o da a acquisi ion and p ocessing, such as: planning, came a calib a ion, 3D econs uc ion and applica ions; and Remondino (2011) p o ides an ex ensi e e iew o de ices and so wa e o eco ding and modelling o ien ed owa ds cul u al he i age. La ely, Kou soudis e al. (2014) p oposed a PG sys em combining UAV and e es ial pic u es and compa ed he esul ing econs uc ion wi h ha ob ained wi h TLS, epo ing p omising esul s. In cons as , Xu e al. (2014) ac ually combined 3D da a om TLS and a UAV-moun ed came a o he econs uc ion o a his o ical monumen in Fujian, China. TLS poin clouds a e used o model he açades and pho og amme ic in o ma ion is used o comple e he oo a ea (no comple ely isible by he TLS sys em loca ed on he g ound). Despi e he g ea e e o in es ed in o in es iga ing he po en ial o TLS and PG o he su ey o his o ic monumen s, we ound ha u he s udies we e s ill necessa y o compa e he quali y o he da a deli e ed by hese echnologies in he speci ic bu common case o s one walls eco ding. In pa icula , he eam wished o assess whe he UAV-based pho og amme y could be used as a eliable means o ob ain su icien ly accu a e ex u ed 3D models o mason y walls o hei su ey and main enance, as such solu ion would conside ably educe he need o and he e o e he cos and sa e y isks associa ed wi h adi ional solu ions o access p o ision (e.g. sca olding, abseiling). 3. Me hodology The compa ison o 3D eali y cap u e echnology was conduc ed expe imen ally in he ield using an ac ual his o ic monumen as es si e. The s udy conside ed: wo e es ial lase scanne s employing di e en measu emen p inciples; a hand-held/pole-moun ed single-came a PG sys em; and a UAV-moun ed single- came a PG sys em. This sec ion i s p esen s he cha ac e is ics and se ings o he es ed sys ems (Sec ion 3.1). The es si e and he easons o i s selec ion a e desc ibed in Sec ion 3.2. Finally, he c i e ia de ined o compa e he pe o mances o he di e en sys ems a e p esen ed in Sec ion 3.3. 3.1. Sys ems The wo TLS de ices conside ed a e he Leica ScanS a ion P40 and he Fa o Focus 3D. The Leica ScanS a ion P40 (he ea e TLS-1) is he la es ime-o - ligh (TOF) scanne p oduced by Leica. I can scan objec s up o 270m away, wi h an accu acy o 3mm a 16 Vale o, Fo s e , Bosche, Wilson, Leslie, 2016. This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0) EDITORIAL UNIVERSITAT POLITÈCNICA DE VALÈNCIA 50m and up o 1 million poin s a e eco ded pe second (Leica Geosys ems 2015). In con as , he Fa o Focus 3D (TLS-2) is a phase-based scanne . I s ad e ised pe o mance (Fa o 2011) cha ac e is ics show ha his scanne can measu e dis ances up o 120m, wi h sligh ly lowe bu s ill good single poin p ecision (2mm a 25m) and can scan up o 1 million poin s pe second. An addi ional impo an ad an age o he Fa o Focus 3D scanne is ha i is signi ican ly smalle and ligh e weigh han scanne s such as he Leica ScanS a ion P40, which makes i mo e po able. The i s single-came a PG sys em conside ed (he ea e PG-1) is a digi al single-lens e lex (DSLR) Nikon D810 came a equipped wi h a 14mm lens ha is ope a ed manually hand-held o moun ed on an ex endable pole. This came a can acqui e 36 megapixel (MPx) pic u es and he sys em allows he acquisi ion o high-quali y da a wi h he came a aised a heigh s up o 5m ( he pole heigh ). The second single-came a pho og amme ic sys em conside ed (he ea e PG-2) is a DSLR Sony -7R came a equipped wi h a 35mm lens ha is moun ed on a UAV and ope a ed emo ely. This came a can also ob ain 36MPx pic u es. The UAV on which he came a is moun ed is an Asc ec Falcon 8 Mul i- o o ha can ly a signi ican dis ances and heigh s – al hough he pilo should con inuously main ain eye con ac , and Ci il A ia ion Au ho i y egula ions ac ually p e en lying such UAVs a heigh s abo e 120 me es. 3.2. Tes Si e The Eas ga den o he medie al C aigmilla Cas le, in Edinbu gh, Sco land, was selec ed as es si e. This cas le was buil in he 14 h cen u y and is cu en ly managed and main ained by His o ic En i onmen Sco land. Wi hin he Eas ga den, he ampa wall ( acing Eas ; see Figu e 1) was selec ed o he ollowing easons: 1. Complexi y: The a ea o wall was selec ed due o i s ela i e complexi y and mo e speci ically: he andom na u e o he s onewo k; a ia ion in wid h and dep h o mo a join ; plana and cu ed su aces; s one soiling and associa e colou and ex u e a ia ion. 2. Su icien heigh : I is impo an o s udy he pe o mance o he chosen echnologies a inc easing heigh s; he ampa wall is app oxima ely 10m high, which seemed su icien o challenge ce ain echnologies, in pa icula he g ound-based TLS de ices. 3. Wind p o ec ion: The UAV can only be ope a ed in condi ions wi h low wind ( o p e en collision wi h he monumen and ensu e good quali y pic u es). To enhance he chances o being able o ope a ed he UAV on he selec ed day o da a acquisi ion, he Eas - acing wall was ound bes due o he p edominan ly wes e n winds in Edinbu gh; 4. Accessibili y: The e should no be ees o o he occlusions p e en ing he sa e use o he UAV and occluding he su eyed scene. The Eas -ga de n does no include any ees. 3.3. Pe o mance Assessmen C i e ia The pe o mances o he ou 3D eali y cap u e sys ems o he eco ding o he 3D geome y o s one walls we e assessed and compa ed using se e al quan i a i e and quali a i e c i e ia. Da a Comple eness: his is measu ed by he densi y and uni o mi y o da a ac oss he acqui ed su ace. E iciency: his is measused by he ime equi ed o acqui ing and p e-p ocessing he da a o ob ain a uni ied geo- e e enced dense 3D poin cloud. Da a Accu acy and P ecision: aking he mos accu a e and p ecise TLS da a as ‘g ound u h’, da a accu acy and p ecision is measu ed by he ‘dis ance’ o he o he da ase s o his g ound- u h da ase . I is p oposed o measu e he ‘dis ance’ be ween wo poin clouds using he Hausdo dis ance me ic (Hu enloche e al. 1993). Fo his, he olume con aining he wall da a is di ided in oxels (i.e cuboids); we use oxels o size 1cm x 1cm x 10cm. The la ge oxel side co esponds o he dep h (pe pendicula o he wall plane); i is se la ge o ensu e ma ching poin s a e ound despi e possible global misalignmen s. No e ha he Hausdo dis ance di e s om he s anda d poin - o-poin dis ance ha simply conside s he closes poin s in he wo poin clouds. In cons as , he Hausdo dis ance is based on he s udy o minimal and maximal dis ances be ween poin clouds wi hin each oxel, which makes i mo e sensi i e o noisy da a. In e es ingly, i also makes i sensi i e o a ia ions in he uni o mi y o he cloud poin densi y. 4. Da a Acquisi ion and Recons uc ion S ages A su ey ne wo k, composed o h ee geo- e e enced ie poin s se wi h 150mm ci cula a ge s, was es ablished wi hin he Eas ga den si e o geo- e e encing he TLS da a. Un o una ely, such su ey ne wo k canno be used o he geo- e e encing o PG da a. This is due o he inabili y o he su eyo s (and he echnologies employed) o simul aneously ocus he came a on he wall and he ie poin a ge s loca ed in he ga den. The e o e, a second su ey ne wo k based on a di e en se o geo- e e enced a ge s had o be es ablished wi h 17cm checke boa d pa e ns ixed o he walls. Bo h su ey ne wo ks a e shown in Figu e 1. All measu emen s o he es ablishmen o he ne wo ks we e conduc ed using a Leica MS50 mul is a ion. To ensu e a ull co e age o he ga den, h ee lase scans we e acqui ed wi h TLS-1, wi h a esolu ion o 1.6mm a a ange o 10m. The h ee scans we e hen aligned using he TLS su ey ne wo k, esul ing in a uni ied clean poin cloud con aining 374 million poin s, including 180 million poin s om he ampa wall. The same p ocedu e was applied o he da a acquisi ion using TLS-2, wi h h ee scans acqui ed a simila loca ions and wi h a esolu ion o 3mm a 10m. The inal uni ied poin cloud con ained 87 million poin s o e all, and 51 million poin s o he ampa wall sec ion. In his case, colou in o ma ion has also been acqui ed using he Fa o Focus 3D in e nal came a. 17 COMPARISON OF 3D REALITY CAPTURE TECHNOLOGIES FOR THE SURVEY OF STONE WALLS This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0) EDITORIAL UNIVERSITAT POLITÈCNICA DE VALÈNCIA Figu e 1: Poin cloud o he Eas ga den o C aigmilla Cas le. The ampa wall is ma ked in ed. The TLS su ey ne wo k is illus a ed in ed, and he PG ne wo k in yellow. Using he PG-1 sys em, 260 pic u es we e aken om many loca ions a ound he ga den and wi h di e en o ien a ion. The dis ance o he came a o he walls was kep o app oxima ely 5m, wi h a ew addi ional pic u es aken a u he dis ances. While some o he images we e acqui ed wi h he came a hand-held, o he images (speci ically o he ampa wall) we e acqui ed wi h he came a moun ed on he pole and igge ed using a mobile phone app. An o e lap o 40-50% be ween adjacen pho os was aimed o . The 3D econs uc ion was conduc ed using Agiso Pho oScan .1.1.6, deli e ing a dense poin cloud con aining 79 million poin s o e all, including 51 million poin s o he ampa wall sec ion. The PG-1 poin cloud was scaled and geo- e e encing using he PG su ey ne wo k. Using he PG-2 sys em, 460 pic u es we e acqui ed a simila dis ances o he walls. The pic u es we e hen p ocessed using he exac same p ocedu es as o he PG-1 sys em, deli e ing a geo- e e enced and scaled dense poin cloud con aining 34 million poin s o e all, including 20 million poin s o he ampa wall sec ion. The econs uc ion p ocess used o bo h pho og amme ic econs uc ions (PG-1 and PG-2) in eg a es he op imisa ion o he came a in e nal calib a ion, and his p ocess is ca ied ou using as ini ial alues hose p o ided by he manu ac u e s (EXIF in o ma ion). The poin clouds co esponding o he ampa wall o all hese econs uc ions a e shown in Figu e 2. No e ha he colou in o ma ion was acqui ed wi h he TLS-2 sys em bu no TLS-1 sys em. None heless, colou in o ma ion appea s o signi ican ly be e quali y o he PG sys ems. 5. Resul s 5.1. Da a P ecision and Accu acy Fo he epo ing and analysis o he esul s on he accu acy and p ecision o he sys ems, we ocus on h ee a eas o he ampa wall ha we ound we e ep esen a i e in highligh ing he s eng hs and limi a ions o he di e en su ey echnologies ega ding accu acy and p ecision. These a eas a e shown in Figu e 3. A ea 1 is an uppe pa o he wall, a 5 me e s heigh and i s dimensions a e 4m x 2m. A ea 2 is loca ed jus below a human heigh and has simila dimensions. Finally, A ea 3 co e s a e ical egion, wi h dimensions 2m x 6m. a) b) c) d) Figu e 2: Wall econs uc ion o he employed sys ems. (a) TLS-1, (b) TLS-2, (c) PG-1, (d) PG-2. Figu e 3: The h ee a eas wi hin he Eas ampa wall ha a e used o he assessmen o pe o mance in e ms o accu acy and p ecision. As p e iously men ioned, all he da ase s we e geo- e e enced by means o su ey ne wo ks. Gi en he ma ches be ween he ie poin s in he su ey ne wo ks and he co esponding poin s in he acqui ed poin clouds, he geo- e e encing e o s a e calcula ed o all de ices and summa ised in Table 1. These esul s, oge he wi h he be e speci ica ions o he TLS-1 de ice in compa ison wi h TLS-2, lead o he conclusion ha TLS-1 indeed p o ides he da a wi h he highes quali y o e all and so can be used as g ound u h agains which all o he da ase s can be compa ed. 18 Vale o, Fo s e , Bosche, Wilson, Leslie, 2016. This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0) EDITORIAL UNIVERSITAT POLITÈCNICA DE VALÈNCIA Table 1: Mean esidual e o s o he geo- e e encing o he di e en de ices. De ice Value (inch) TLS-1 1.53 TLS-2 3.73 PG-1 19.8 PG-2 18.1 Figu e 4 illus a es he esul s o he Hausdo dis ance calcula ions o he h ee selec ed a eas o he ampa wall, and Table 1 summa izes he mean and s anda d de ia ions o he Hausdo dis ance o each case. As can be seen, he Hausdo dis ances a e smalle o he Fa o TLS wi h a mean below 5mm in bo h A eas 1 and 2. The mean dis ance o he PG-1 poin cloud is only sligh ly highe and is hus ema kably good. In con as , he esul s a e disappoin ing o he PG-2 poin cloud, wi h a mean be ween 10 and 20mm. The esul s a e pa icula ly poo o A ea 1. Bu , we iden i ied ha his was due o he insu icien amoun o o e lap be ween neighbou ing pic u es acqui ed by he PG-2 sys em. The easons o he poo pe o mance o PG-2 a e discussed u he in Sec ion 5.4. A ea 3 was selec ed o assess he a ia ion o he accu acy and p ecision wi h heigh o he di e en sys ems. I is in e es ing o no e he ela i e uni o mi y o he Hausdo dis ances along he heigh o he wall o he PG-2 da a which highligh s he ad an age o UAV in e ms o access, he eby ensu ing consis en esul s. Fo he pole-moun ed PG-1 sys em, he dis ances inc ease mo e signi ican ly a he bo om, because ewe images we e aken o he lowe pa s – he ope a o was ocused on he co e age o he uppe pa o he wall ha is less accessible and did no ealise ha an insu icien numbe o images would be aken o he lowe pa (e.g. aking pic u es while on hei knees). Howe e , he heigh o he wall (10m) seemed no o ha e been qui e su icien o show a simila pa e n in he op pa o he wall esul ing om he limi ed ex ension o he pole. This sugges s ha e en wi h a 5m pole, good-quali y da a can s ill be ob ained up o almos wice ha heigh . O e all, hese esul s con i m ha TLS de ices p o ide mo e p ecise and accu a e esul s han pho og amme ic sys ems. Bu , he cheape pole-moun ed PG-1 sys em demons a ed compa a i ely good esul s, meaning ha such a sys em can ce ainly be conside ed as a eliable al e na i e o he su ey o s one walls. Table 2: Mean (µH) and s anda d de ia ion (σH) o he Hausdo dis ance o he TLS-2, PG-1 and PG-2 poin clouds agains he TLS-1 poin cloud, o he h ee selec ed A eas o he ampa wall. De ice A ea 1 (mm) A ea 2 (mm) A ea 3 (mm) µH σH µ H σH µ H σH TLS-2 3.67 1.06 4.29 0.99 5.23 1.66 PG-1 4.36 1.60 5.09 1.95 6.14 2.8 PG-2 17.62 10.18 10.79 5.62 7.76 3.04 5.2. Da a Comple eness Pe o mance canno be solely judged on da a p ecision and accu acy. Ano he impo an c i e ion is da a comple eness, ha assesses da a poin densi y and i s uni o mi y. Table 3 summa izes he poin densi y in he poin clouds ob ained o he en i e ampa wall by means o he ou di e en sys ems. Poin densi y is calcula ed by p ojec ing he econs uc ed poin s on he wall plane and hen calcula ing he densi y o poin s wi hin ha plane. As can be seen, he densi y is highe o he lase scanne s, bu again we no e ha he di e ence be ween he TLS-2 and PG-1 sys ems is eally no ha signi ican , which u he con i ms he po en ial o PG-1 sys em as an al e na i e 3D eali y cap u e echnology. Table 3: The maximum densi y (MaxD), mean densi y (µD) and s anda d de ia ion (σD) o he poin clouds o he en i e ampa wall ob ained by he ou eali y cap u e sys ems. De ice MaxD (p s/cm 2 ) µD (p s/cm 2 ) σD (p s/cm 2 ) TLS-1 164 75.94 11.67 TLS-2 49 22.99 3.88 PG-1 31 16.09 3.05 PG-2 23 7.54 2.29 5.3. P ocess E iciency An addi ional c i e ion o compa ing pe o mance be ween TLS and PG sys ems is e iciency. This is de ined as he ime equi ed o acqui ing and p ocessing he da a o ob ain a uni ied geo- e e enced dense 3D poin cloud. Acquisi ion ime comp ises he posi ioning o he de ices a ound he ga den and he acquisi ion o poin s clouds ( o TLS de ices) o pic u es ( o PG sys ems). P ocessing includes da a ans e , egis a ion/geo- e e encing, colou isa ion and da a cleaning ( emo ing spu ious da a). Table 5 shows he app oxima e ime eco ded o each ask o each su ey sys em. No e ha he p ocessing ope a ions ha e all been unde aken on he same compu e (i7 3.60GHz p ocesso and 12 GB RAM). Rega ding he acquisi ion s age, he scanning ime appea s signi ican ly sho e o he TLS-1 han o he TLS-2. Bu , his is essen ially due o he ac ha no colou in o ma ion was acqui ed du ing he TLS-1 scanning. I colou acquisi ion had been conduc ed, he acquisi ion imes would likely ha e been simila . An impo an di e ence is no iceable be ween PG-1 and PG-2. This is due o wo ac o s. Fi s , he longe ocal leng h selec ed o he PG 2 sys em led o he need o acqui e many mo e images wi h he PG-2 sys em ( o ensu e su icien o e lap be ween neighbou ing images). Then, he UAV could only wo k o 10-minu e pe iods a a ime, a e which he UAV had o be landed o change i s ba e y be o e esuming wo ks. Combined, hese wo ac o s esul ed in a signi ican ly longe acquisi ion ime han expec ed o he PG-2 sys em. The analysis o he pe o mance o he PG-2 sys em is u he discussed in Sec ion 5.4. 19 COMPARISON OF 3D REALITY CAPTURE TECHNOLOGIES FOR THE SURVEY OF STONE WALLS This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0) EDITORIAL UNIVERSITAT POLITÈCNICA DE VALÈNCIA TLS-2 PG-1 PG-2 A ea 1 A ea 2 A ea 3 Figu e 4: Hausdo dis ance be ween Leica ScanS a ion P40 poin cloud and he poin clouds o Fa o (le ), Nikon (middle) and Sony ( igh ) o he h ee selec ed A eas o he ampa wall. P ocessing imes a e simila o all sys ems, despi e he ac ha hey equi e di e en p ocessing s eps. The main obse a ion is ha p ocessing imes a e in his case s udy o en mo e han double he acquisi ion imes, despi e he use o a high-pe o mance compu e . Table 4: Acquisi ion and p e-p ocessing imes o TLS and PG sys ems. Acquisi ion P ocessing To al TLS-1 40’ 3 h 3h 40’ TLS-2 1h 30’ 2 h 30’ 4 h PG-1 30’ 3 h 15’ 3 h 45’ PG-2 1h 30’ 3 h 15’ 4 h 45’ 5.4. Discussion o he Pe o mance o he UAV-based PG Sys em (PG-2) Reasonably good quali y esul s we e achie ed wi h he PG-1 sys em, sugges ing ha in such con ex s pho og amme ic sys ems can be conside ed as iable al e na i es o TLS. I was hoped ha simila esul s would also be achie ed wi h he PG-2 sys em ha would ha e demons a ed i s addi ional ad an age in e ms o access p o ision. Ye , he esul s achie ed wi h he PG- 2 sys em we e compa a i ely e y poo and disappoin ing o e all. I mus howe e be quali ied ha , in some a eas o he wall, he esul s ob ained wi h he PG-2 sys em we e as good (i no be e ) han hose ob ained wi h he PG-1 sys em. This sec ion aims o discuss he easons o hese poo esul s obse ed in ce ain a eas, highligh ing he ac o s ha con ibu ed o ge ing hose poo esul s, and how hey could in ac ha e been ob ia ed. This shall demons a e ha UAV- based PG sys ems do ha e g ea po en ial o he su ey o s one walls, as long as he iden i ied ac o s a e adequa ely managed. The local a iabili y in he quali y o he econs uc ion ob ained wi h he PG-2 has highligh ed ha (1) came a loca ions, and (2) came a p ope ies and se ings can sign ican ly impac he pe o mance o hose sys ems. PG-1 and PG-2 ac ually used di e en came a se ups and came a posi ioning me hods, which we belie e al oge he explain he poo e esul s achie ed by PG-2. Fi s ly, Table 5 shows ha he came a se ings o PG-1 and PG-2 di e ed in se e al ways. The PG-2 sys em did no use en i ely manual se ings, which led o images wi h a ying ISO and F-s op alues. Bu , an impo an di e ence is no iced in he lens’s ocal leng h. The ocal leng h o PG-2 is h ee imes la ge han ha o PG-1, which implies ha he wall a ea co e ed by each pic u e aken by he PG-2 sys em was signi ican ly smalle . The g ound sampling dis ance (GSD) o each sys em can be ob ained om he pa ame e s p esen ed in Table 5. Fo PG-1, GSD=2mm; o PG-2, GSD=0.5mm. To ensu e a su icien amoun o o e lap be ween pic u es, a signi ican ly la ge numbe o images was hus necessa y. Al hough wice mo e pic u es we e indeed acqui ed using he PG-2 sys em compa ed o PG-1, his numbe appea s o s ill ha e been insu icien . Indeed, he poo quali y o he econs uc ions achie ed in he local a eas o he ampa wall A ea 1 and A ea 2 isible in Figu e 4 appea s o di ec ly co ela e wi h a lowe o e lap be ween images acqui ed wi h he UAV in hese a eas, as can be seen in Figu e 5. 20 Vale o, Fo s e , Bosche, Wilson, Leslie, 2016. This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0) EDITORIAL UNIVERSITAT POLITÈCNICA DE VALÈNCIA Table 5: P ope ies and se ings o he came as used in he wo PG sys ems. Se ings PG-1 PG-2 Size (px) 7360 x 4912 7360 x 4912 Senso (mm) 35.9 x 24 35.9 x 24 Focal leng h (mm) 14 35 F-s op /8 4- /8 ISO 250 100-400 Shu e (s) 1/125 1/500 Numbe o pic u es 260 460 A ea o wall co e ed by he came a a 5m (m) 15.5 x 10.3 4 x 2.7 Figu e 5: Came a loca ions (in blue) o he acquisi ion o da a wi h he PG-2 sys em.The h ee a eas o he wall used o he pe o mance assessmen a e shown in ed. The PG 2 sys em ope a o s may ha e expec ed ha , by acqui ing mo e images wi h a lens ha ing a longe ocal leng h, be e esul s would be p oduced. Bu , hey unde es ima ed he ull impac o a longe ocal leng h on he numbe o images needed o be acqui ed. In his case, i seems ha hey should p obably ha e acqui ed wice mo e pic u es han hey ac ually ha e (which would ha e u he nega i ely in luenced he acquisi ion ime). Na u ally, o he ac o s may ha e in luenced he esul s, such as he a iable F-s op and ISO alues. Bu , we ound he PG-2 pic u es o be o quali y simila o ha o he PG-1 sys em. This sugges s ha , hanks o hei mode n gimbals, UAVs can p o ide a eliable pla o m o he acquisi ion o images o pho og amme ic pu poses. O e all, we belie e ha signi ican ly be e esul s (a leas as good as PG-1) would ha e been ob ained by using a lens wi h sho ocal leng h (14-16mm) on he UAV-moun ed came a (PG-2). A mo e gene al lesson d awn om his wo k is ha ca e ul planning should be conside ed o selec he app op ia e came a se ings and pic u e densi y (and loca ions) o ensu e accu a e and eliable pho og amme ic econs uc ion. This p ocess may be e e ed o as: Planning o Pho og amme y (P4P). 6. Conclusions Te es ial lase scanning and pho og amme y a e inc easingly used o building su eying, p o iding dense ex u ed 3D poin clouds (and meshes). In e ms o geome y, da a om TLS was ound pa icula ly accu a e. While his was expec ed (since i had been shown by p e ious esea che s), in he case s udy conside ed he e he Fa o Focus 3D scanne poin clouds we e close o he Leica ScanS a ion P40 ones. This simila i y is in e es ing, al hough i mus be emembe ed ha hese esul s a e ob ained a a he sho anges (maximum 10m) and he pe o mance o he Fa o scanne would be expec ed o de e io a e as e a la ge dis ances. While TLS clea ly p o ides good geome ic da a o gene a e accu a e and aluable 3D models, i also has h ee impo an limi a ions. Fi s , TLS de ices emain ela i ely expensi e ( ens o housands o pounds pe uni ). Also, he came as embedded wi hin he scanne s do no p oduce good quali y colou in o ma ion compa ed o wha is achie ed wi h common DSLR came as. And inally, bu s ill e y impo an ly, cu en TLS de ices ha e o be ope a ed om s able posi ions, which educes hei mobili y and he ange o con ex s wi hin which hey can be eliably employed. In con as , mode n digi al came as a e ela i ely cheap and e y po able, making hem sui able in a wide ange o con ex s. Howe e , hei main limi a ion is hei a iable pe o mance depending on he le el o ex u e in he scene being econs uc ed. None heless, we ha e shown ha scenes like s one walls p esen g ea ex u es o he applica ion o pho og amme y. In such con ex s, pho og amme y can cons i u e a ealis ic al e na i e o TLS. Moun ing a came a on a UAV can u he sol e access issues. A cop e - ype UAV can ly a ound close o buildings and ake pic u es om di e en iewpoin s wi hou he need o any addi ional in as uc u e (e.g. sca olding). This heo e ically ex ends he applicabili y o pho og amme y o signi ican ly mo e con ex s (e.g. wi hou occluded a eas, as was expe ienced wi h TLS). Un o una ely, his esea ch was no able o ully demons a e his. Howe e , he analysis o he ac o s ha led o he disappoin ing esul s ob ained wi h he UAV-moun ed PG sys em we e ‘simply’ an inapp op ia e selec ion o lens and a co espondingly insu icien numbe o images acqui ed. The esul s showed he need o pho og amme ic su ey eams o ha e a good unde s anding o he pho og amme ic p ocess o es ablish e ec i e da a acquisi ion p ocedu es. This sugges s he need o es ablish obu s me hods o Planning o Pho og amme y (P4P), which should be he ocus o u u e in es iga ions. Bu , he esul s epo ed he e should no de e p o essionals om conside ing his o he wise aluable echnology ha could signi ican ly educe he cos and sa e y isks associa ed wi h access p o ision, in compa ison wi h cu en solu ions used o manual su ey as well as he use o TLS sys ems. Beyond he echnical conclusions p o ided, his wo k also highligh s ha mixed adop ion o echnologies o speci ic echnologies u ilised o su ey a e po en ially equi ed. Thei selec ion will depend upon weigh ed conside a ions o da a p ecision, acquisi ion and p ocessing e iciency, and heal h and sa e y g ounds. The wo k shows ha i is impo an ha su eyo s a emp o be e unde s and he p os and cons o echnologies adop ed and ealise ha some will pe o m be e han o he s in di e en con ex s. I s esses ha be e dialogue be ween adi ional su eyo s and a chi ec s and specialis echnologically-o ien ed su ey ope a i es is equi ed so ha be e unde s anding o expec a ions can be achie ed by bo h pa ies. 21 COMPARISON OF 3D REALITY CAPTURE TECHNOLOGIES FOR THE SURVEY OF STONE WALLS This wo k is licensed unde a C ea i e Commons 4.0 In e na ional License (CC BY-NC-ND 4.0) EDITORIAL UNIVERSITAT POLITÈCNICA DE VALÈNCIA Acknowledgemen s The w i e s a e g a e ul o His o ic En i onmen Sco land (HES) o unding his p ojec , and Cybe hawk Inno a ions L d. o conduc ing he UAV-based su ey. Re e ences BAY H., TUYTELAARS, T and VAN GOOL, L., 2006. SURF: Speeded Up Robus Fea u es, P oceedings o he 9 h Eu opean Con e ence on Compu e Vision, Pa I, G az, Aus ia, pp. 404-417. CAPELLINI, V., STEFANI, C., NONY, N. and DE LUCA, L., 2012 Su eying Mason y S uc u es by Seman ically En iched 2.5D Tex u es: A New App oach. In M. IOANNIDES, D. FRITSCH, J. LEISSNER, R. DAVIES, F. REMONDINO and R. CAFFO, eds, P og ess in Cul u al He i age P ese a ion, Vol. 7616 o Lec u e No es in Compu e Science, Sp inge Be lin Heidelbe g, pp. 729–737. DOI; 10.1007/978-3-642-34234-9_77 CARDACI, A., ROBERTI, G. M. and VERSACI, A., 2011. F om he Con inuous o he Disc e e Model: a Lase Scanning Applica ion o Conse a ion P ojec s. In P oceedings o he Cong ess o he In e na ional Socie y o Pho og amme y and Remo e Sensing (ISPRS) 2011, T en o, I aly, pp. 437–444. FARO, 2011. Fa o Focus 3D 120 Speci ica ions. A ailable: h p://www. a o.com/p oduc s/3d-su eying/lase -scanne - a o- ocus-3d [1/26, 2016]. HISTORIC SCOTLAND, 2011. Sco ish T adi ional Building Skills. Audi Scoping Exe cise. Technical epo , His o ic Sco land. HUTTENLOCHER, D., KLANDERMAN, G. and RUCKLIDGE, W., 1993. Compa ing Images Using he Hausdo Dis ance. IEEE T ansac ions on Pa e n Analysis and Machine In elligence 15(9), pp. 850–863. DOI: 10.1109/34.232073. KOUTSOUDIS, A., VIDMAR, B., IOANNAKIS, G., ARNAOUTOGLOU, F., PAVLIDIS, G. and CHAMZAS, C., 2014. Mul i- Image 3D Recons uc ion Da a E alua ion. Jou nal o Cul u al He i age 15, pp. 73–79. DOI: doi:10.1016/j.culhe .2012.12.003. LEICA GEOSYSTEMS, 2015. Leica ScanS a ion P30/P40 Speci ica ions. A ailable: h p://w3.leica- geosys ems.com/downloads123/hds/hds/gene al/b ochu es-da ashee /Leica_ScanS a ion_P30- P40_Plan _DS_en.pd [1/26, 2016]. LERMA, J. and MUIR, C., 2014. E alua ing he 3D Documen a ion o an Ea ly Ch is ian Up igh S one wi h Ca ings om Sco land wi h Mul iples Images. Jou nal o A chaeological Science 46, pp.311–318. DOI: doi:10.1016/j.jas.2014.02.026. LOWE, D., 1999. Objec Recogni ion om Local Scale-In a ian Fea u es. In P oceedings o he Se en h In e na ional Con e ence on Compu e Vision, 1999, Ke ky a, G eece, Vol. 2, pp. 1150–1157. MUELLER, G., ROBERTSON, J., LEADBETTER, C., LAING, N., MCMENEMY, M. and KYRIAKOU, A., 2014. Sco ish House Condi ions Su ey 2013. Key Findings. Technical epo , Di ec o a e o Housing, Regene a ion and Wel a e. Sco ish Go e nmen . NETTLEY, A., ANDERSON, K., DESILVEY, C. and CASELDINE, C., 2011. Using Te es ial Lase Scanning and LIDAR Da a o Pho o-Realis ic Visualiza ion o Clima e Impac s a He i age Si es. In e na ional A chi es o Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sciences, 38, pp. 223–229. DOI: 10.5194/isp sa chi es-XXXVIII-5-W16- 223-2011. OGLEBY, C. L., 1995. Ad ances in he Digi al Reco ding o Cul u al Monumen s. ISPRS Jou nal o Pho og amme y and Remo e Sensing 50(3), 8 – 19. DOI:10.1016/0924-2716(95)91286-S. PUSCHEL, H., SAUERBIER, M. and EISENBEISS, H., 2008. A 3D Model o Cas le Landenbe g (CH) om Combined Pho og amme ic P ocessing o Te es ial and UAV-Based Images. In e na ional A chi es o he Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sciences, 37, pp. 93–98. RASZTOVITS, S. and DORNINGER, P., 2013. Compa ison o 3d Recons uc ion Se ices and Te es ial Lase Scanning o Cul u al He i age Documen a ion. XVI In e na ional CIPA Symposium, 2013, S asbou g, F ance, Vol. 1, pp. 513–518. REMONDINO, F., 2011. He i age Reco ding and 3D Modeling wi h Pho og amme y and 3D Scanning. Remo e Sensing 3, pp. 1104–1138. DOI: 10.3390/ s3061104. REMONDINO, F., BARAZZETI, L., NEX, F., SCAIONI, M. and SARAZZI, D., 2011. UAV Pho og amme y o Mapping and 3D Modeling - Cu en S a us and Fu u e Pe spec i es. In e na ional A chi es o he Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sciences, 38, pp. 25–31. SCHERMERHORN, W., 1940. D a o an In e na ional Bibliog aphy o Pho og amme y, Pho og amme ia 3(3), pp. 116 – 119. DOI: 10.1016/S0031-8663(40)80036-7. 22