Development of a New Low-Cost Indoor Mapping System – System Design, System Calibration and First Results

DEVELOPMENT OF A NEW LOW-COST INDOOR MAPPING SYSTEM – SYSTEM DESIGN, SYSTEM CALIBRATION AND FIRST RESULTS T. P. Ke s en a*, D. S allmann a, F. Tschi schwi z a a Ha enCi y Uni e si y Hambu g, Pho og amme y & Lase Scanning Lab, Übe seeallee 16, D-20457 Hambu g, Ge many - (Thomas.Ke s en, Di k.S allmann, Felix.Tschi schwi z)@hcu-hambu g.de Commission V, WG V/1 KEY WORDS: 3D, came a, calib a ion, modelling, spa se poin cloud, pano ama, econs uc ion ABSTRACT: Fo mapping o building in e io s a ious 2D and 3D indoo su eying sys ems a e a ailable oday. These sys ems essen ially di e om each o he by p ice and accu acy as well as by he e o equi ed o ieldwo k and pos -p ocessing. The Labo a o y o Pho og amme y & Lase Scanning o Ha enCi y Uni e si y (HCU) Hambu g has de eloped, as pa o an indus ial p ojec , a low- cos indoo mapping sys em, which enables sys ema ic in en o y mapping o in e io acili ies wi h low s a ing equi emen s and educed, measu able expendi u e o ime and e o . The modelling and e alua ion o he eco ded da a ake place la e in he o ice. The indoo mapping sys em o HCU Hambu g consis s o he ollowing componen s: lase ange inde , pano ama head (pan- il -uni ), single-boa d compu e (Raspbe y Pi) wi h digi al came a and ba e y powe supply. The came a is p e-calib a ed in a pho og amme ic es ield unde labo a o y condi ions. Howe e , emaining sys ema ic image e o s a e co ec ed simul aneously wi hin he gene a ion o he pano ama image. Due o cos easons he came a and lase ange inde a e no coaxially a anged on he pano ama head. The e o e, eccen ici y and alignmen o he lase ange inde agains he came a mus be de e mined in a sys em calib a ion. Fo he e i ica ion o he sys em accu acy and he sys em calib a ion, he lase poin s we e de e mined om measu emen s wi h o al s a ions. The di e ences o he e e ence we e 4-5mm o indi idual coo dina es. 1. INTRODUCTION Digi al building models ep esen he basis o planning, cons uc ion and eno a ion as well as o he managemen o buildings. In his con ex Building In o ma ion Modelling (BIM) desc ibes a me hod o op imized h ee-dimensional planning, execu ion o cons uc ion wo k and managemen o buildings by app op ia e so wa e so ha all ele an building da a a e digi ally eco ded, combined and in e connec ed. BIM applies bo h in he building indus y o he building, planning and execu ion o cons uc ion wo k (a chi ec u e, enginee ing, building se ices), and in he acili y managemen sec o . P ima ily, he a eas o buildings in he building and eal es a e economy ep esen he basis o p ojec planning, cos ing and income planning as well as o cons uc ion inancing. Since building a eas a e also used o he appo ionmen o ope a ing cos , incomple e, misleading o alse da a abou building a eas can ha e ex ensi e, unwan ed consequences (Kalusche 2011). Fo 90% o eal es a e in en o y a ea calcula ions a e no co ec (P esseBox 2005, Wagne 2015). Thus he eal es a e indus y uses inco ec da a abou building s ock in many cases, i.e. eliable cu en da a is o en no a ailable a all. In o de o measu e cu en a eas o p ope ies, di e en mode n measu emen me hods such as e es ial lase scanning (Ke s en e al. 2005, Ke s en & Linds aed 2012), digi al pho og amme y (Ke s en e al. 2004) and achyme y a e a ailable, all o which wo k sa is ac o ily o 3D eco ding o a chi ec u al objec s and which a e es ablished in he ma ke . Howe e , hese p ocedu es and sys ems equi e a ce ain le el o expe knowledge, unde pinned by high acquisi ion and educa ion cos s. A lase - pho og amme ic imaging sys em using pano ama pho og aphy * Co esponding au ho and lase anging in he opposi e axis o he came a was p esen ed by Clauß (2011) and Clauß (2012). He ing (2012) showed in his in es iga ions in o ins umen and model accu acy, ha his lase - pho og amme ic imaging sys em is able o ensu e measu emen s up o an accu acy o 2cm. Fangi (2007) p esen ed an app oach whe e se e al sphe ical pano ama pho og aphs we e used o he 3D modelling o he in e io s o wo I alian chu ches. Fu he mo e, so-called dep h senso s ( ime-o - ligh came as) ha e al eady been used o he eco ding o in e io s (Zhu & Donia 2013, Hen y e al. 2014). How seman ic in o ma ion can be au oma ically ex ac ed om poin clouds de i ed om pho og aphy o he in e io o BIM applica ions has been desc ibed by Tamke e al. (2014) and K ispel e al. (2015). The Labo a o y o Pho og amme y & Lase Scanning o he Ha enCi y Uni e si y Hambu g is de eloping a low-cos imaging sys em o in e io s in he con ex o a R&D coope a ion p ojec o he cen al inno a ion p og amme (ZIM) o small and medium-sized en e p ises (SMEs). Pa ne s in his p ojec a e he Ins i u e o Compu e Science (Mul imedia In o ma ion P ocessing) o he Ch is ian-Alb ech s Uni e si y Kiel, Ge many and he company ETU So wa e GmbH om Cologne, Ge many. The goal o his p ojec is o de elop new o signi ican ly imp o e exis ing p oduc s, p ocesses and echnical se ices. In he ollowing he sys em con igu a ion (chap e 2), he came a and sys em calib a ion (chap e 3 and 4), he da a acquisi ion (chap e 5) and he 3D modelling (chap e 6), including i s esul s, will be desc ibed. 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, Volume XLI-B5, 2016 XXIII ISPRS Cong ess, 12–19 July 2016, P ague, Czech Republic This con ibu ion has been pee - e iewed. doi:10.5194/isp sa chi es-XLI-B5-55-2016 55 2. SYSTEM CONFIGURATION The indoo mapping sys em o he HCU Hambu g is designed o cap u e comple e 360° pano ama images o in e io spaces. The o e all sys em consis s o a pano ama head wi h a came a, which is i mly moun ed wi h a lase ange inde on a common pla o m. The lase ange inde poin s in o he di ec ion o he came a and measu es simul aneously a dis ance o a nonspeci ic poin in he in e io o each aken pho o. In o de o ge i s expe iences and acqui e es da a, a p o o ype o he low-cos mapping sys em was buil a HCU Hambu g using he ollowing componen s (Fig. 1): 1. Con ol compu e : Raspbe y Pi Model B 2. Digi al came a: Raspbe y Pi came a module 3. Lase ange inde : Bosch GLM 100 C P o essional 4. Powe supply: LogiLink powe bank 5. Pan- il uni : Man o o mul i-line pano ama head 6. Le elling de ice T ib ach wi h ubula le el 7. T ipod: Man o o ipod The ad an ages o he selec ed ha dwa e a e: (a) The cos s o he combina ion o Raspbe y Pi single-boa d compu e and came a module a e e y a ou able wi h less han 60 EUR. (b) A la ge ange o accesso ies o a ious applica ions al eady exis s o he Raspbe y Pi compu e . (c) Since he compu e can be ope a ed wi h di e en Open-Sou ce 32-Bi ope a ing sys ems, a e y ex ensi e o e ing o so wa e lib a ies is a ailable. (d) The GPIO (gene al pu pose inpu /ou pu ) in e ace o he compu e can simply be used o he con ol o senso s and displays. (e) The Bosch lase ange inde can be comple ely con olled ia Blue oo h and he USB in e ace. ( ) The Man o o mul i-line pano ama head ul ils wo subs an ial unc ions o he gene a ion o sphe ical pano amas: 1. The o a ion abou a second axis ( il ing axis), in o de o ake pho og aphs wi h di e en e ical angles. 2. The mo emen o he came a a ound he en ance pupil o he lens (nodal poin ). Con ol so wa e wi h a simple use in e ace uns on he Raspbe y Pi compu e , which enables acquisi ion o he images and he dis ance measu emen s as well as he egis a ion o he came a axes (ho izon al and e ical angles). The lase ange inde is connec ed o he Raspbe y Pi compu e ia USB in e ace. The Raspbe y Pi compu e is equipped wi h a WiFi USB adap e , which se s up a local WLAN ne wo k, so ha he sys em can be ope a ed by emo e connec ion om an ex e nal pe sonal compu e , e.g. a lap op. 3. CAMERA CALIBRATION To acqui e geome ically co ec in o ma ion om he pano amic images, he digi al came a has been selec ed no only o cos easons bu also by i s op ical cha ac e is ics such as scale, dis o ion and s abili y. As a me hod o he objec i e alida ion o hese c i e ia came a calib a ion in a 3D es ield was selec ed. Fo he geome ical in es iga ions o he came a a he new building loca ion o he HCU Hambu g a p o isional es ield was es ablished (Fig. 2). The es ield has a size o 4.5m (wid h) × 3.3m (heigh ) × 2.0m (dep h). The measu emen s o he con ol poin s we e ca ied ou by a o al s a ion Leica TM30. Fo he adjus men o he geode ic 3D ne wo k he p og am PANDA om GEOTEC was used. The s anda d de ia ions o he adjus ed con ol poin coo dina es we e 0.3mm (XY) in plane and 0.9mm (Z) in dep h. Figu e 2. Tempo a y 3D es ield o HCU Hambu g Fo he came a calib a ion ou Raspbe y Pi came as (RPI 1, 2, 3, 4) wi h no mal angle lens (c = 3.6mm) and one came a (RPI 5) wi h wide angle lens (c = 1.7mm) we e a ailable. The calib a ion p ocedu e was epea ed wi h he di e en came as as indica ed in Tab. 2. The Raspbe y Pi Founda ion cu en ly o e s h ee di e en came a modules. All came a modules ha e he same 5-Mega- Pixel CMOS senso o he ype OmniVision OV5647 wi h 1.4 µm pixel spacing. The echnical speci ica ions o he Raspbe y Pi came a a e summa ized in Tab. 1. The es ed came as a e p obably no iden ically cons uc ed. They di e ou wa dly by Figu e 1. Sys em componen s o he HCU 3D-IMAGER ( .l. . .): Raspbe y Pi compu e and came a, Bosch lase dis ance measu ing ins umen , Man o o pano ama head, USB powe bank and he assembled HCU 3D-IMAGER sys em 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, Volume XLI-B5, 2016 XXIII ISPRS Cong ess, 12–19 July 2016, P ague, Czech Republic This con ibu ion has been pee - e iewed. doi:10.5194/isp sa chi es-XLI-B5-55-2016 56 he layou o he pla e and he insc ip ion o he chips. The cen al componen , he OV5647 is p obably he same o all. Un o una ely, no se ial numbe is used o clea iden i ica ion o each came a module. Came a OmniVision OV5647 Senso CMOS colou , 2592×1944 pix (5 MP) Video 1080p wi h 30 ames/s, H.264 coding Pixel size 1.4µm × 1.4µm Senso size H×V 3.62mm × 2.72mm Colou dep h 10 bi In e ace CSI Con ol Mul i-Media Abs ac ion Laye Lens in eg a ed, ixed ocal leng h & ape u e FoV H×V 54° × 41° (no mal angle) Size H×W×D 25mm × 24mm × 9mm Weigh 3 g Table 1. Technical speci ica ions o he Raspbe y Pi came a Fo came a calib a ion he bundle adjus men p og am Pic an om Techne GmbH was used. The Pic an came a model desc ibes he came a by h ee pa ame e s o in e io o ien a ion (p inciple poin x0, y0 and came a cons an c) and by six addi ional pa ame e s A1…A6 (sys ema ic image e o s), whe eby A1 and A2 model he adial symme ic dis o ion o he lens, A3 and A4 he a ini y and shea ing o he senso chip and A5 and A6 he adial asymme ic and angen ial dis o ion o he lens. The adjus ed nine pa ame e s o he in e io o ien a ion a e summa ized in Table 2 a e calib a ion o i e di e en Raspbe y Pi came as ( ou wi h no mal lens and one wi h wide angle lens). The e ec s o he addi ional pa ame e s on he image coo dina es we e de e mined o he pa ame e g oups (A1, A2), (A3, A4) and (A5, A6) sepa a ely and all oge he . These e ec s a e ep esen ed in Fig. 3 wi h a 11 × 11 g id on he senso chip. The g een g id shows he senso (chip size is 3.63mm × 2.72mm) and he ed g id shows he e ec o he addi ional pa ame e s o images coo dina es wi h 10 imes enla gemen . Fo he calib a ed came as he ollowing esul s a e alid: 1. The e ec o he adial symme ic dis o ion (A1, A2) is app oxima ely +20µm a he co ne s o he senso . The alues a e simila o di e en came as modules (Fig. 4). 2. The e ec o a ini y and shea ing (A3, A4) is usually smalle han 1µm. 3. The e ec o he adial asymme ic and angen ial dis o ion (A5, A6) is usually smalle han 3µm. The adial symme ic dis o ion has he la ges sys ema ic e ec on he image coo dina es. All pa ame e s a e signi ican ly de e mined in he bundle block adjus men . Fu he mo e, he alues show li le a ia ion o he di e en came as. Figu e 3. E ec o he addi ional pa ame e s (A1 … A6) on he image coo dina es The epea ed calib a ions show he ollowing esul s: 1. The came a cons an c a ied only sligh ly and seems o be s able. 2. On he o he hand he posi ion o he p inciple poin shows de ia ions up o 50 µm. 3. The adial symme ic dis o ion (A1, A2) a ied only sligh ly and seems o be ela i e s able (Fig. 4). 4. The o he addi ional pa ame e s (a ini y and shea ing (A3, A4) and adial asymme ic and angen ial dis o ion (A5, A6)) demons a e no clea end. Howe e , hese las wo pa ame e g oups only ha e a ela i ely small e ec on he inal image coo dina es. C M x0 [mm] y0 [mm] c [mm] A1 [E-03] A2 [E-03] A3 [E-05] A4 [E-05] A5 [E-04] A6 [E-05] Cal RPI 1 1 -0.0304 0.0084 3.5864 5.67 -1.53 2.32 1.55 1.60 -2.95 15.04.14 2 0.0185 -0.0048 3.5851 5.74 -1.54 3.78 1.75 -1.98 2.22 26.06.14 3 -0.0230 -0.0001 3.5820 5.59 -1.50 2.90 2.68 1.89 -2.60 12.02.15 4 -0.0233 0.0003 3.5851 5.34 -1.43 48.30 1.09 1.85 -4.50 10.03.15 RPI 2 1 -0.0158 -0.0052 3.5892 5.37 -1.39 43.70 5.38 -1.80 -5.38 26.06.14 2 -0.0058 0.0346 3.5911 5.58 -1.44 46.00 -1.80 -3.64 -2.29 16.02.15 RPI 3 1 0.0096 -0.0371 3.6004 5.50 -1.45 46.60 -0.59 -1.39 -18.50 26.06.14 2 -0.0142 -0.0398 3.5999 5.40 -1.44 -0.48 0.85 0.99 20.00 12.02.15 3 -0.0154 -0.0409 3.6006 5.32 -1.38 44.50 0.49 1.07 15.70 10.03.15 RPI 4 1 0.0137 0.0042 3.5878 5.32 -1.40 46.40 4.82 1.42 -0.06 16.02.15 2 0.0128 0.0039 3.5880 5.27 -1.37 46.50 7.63 1.39 -2.97 18.02.15 RPI 5 1 0.0604 0.0042 1.7421 -77.00 3.66 104.00 17.80 3.17 -15.30 18.02.15 C…came as, M…measu emen s, x0/y0…p inciple poin , c…came a cons an , A1-A6…sys ema ic e o s, Cal… calib a ion da e Table 2. In e io o ien a ion pa ame e se o di e en Raspbe y Pi came as 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, Volume XLI-B5, 2016 XXIII ISPRS Cong ess, 12–19 July 2016, P ague, Czech Republic This con ibu ion has been pee - e iewed. doi:10.5194/isp sa chi es-XLI-B5-55-2016 57 Figu e 4. Radial symme ical dis o ion (A1, A2) o ou di e en Raspbe y Pi came as ( op) and epe i ion o calib a ions o a single came a (bo om) The Raspbe y Pi came a has been compa ed wi h wo digi al SLR came as (Nikon D700 and D90, bo h wi h 20mm lens) and an indus ial came a IDS (wi h au o ocus) in a 3D es ield calib a ion, in o de o ge in o ma ion abou he accu acy po en ial o he low-cos came a. The esul s o he compa ison a e summa ized in Table 3. The image measu ing accu acy o he Raspbe y Pi came a wi h 0.2 pixels (1/5 pixel) is nea ly as good as he SLR came as, while he empi ical accu acy (compa ison o pho og amme ically de e mined poin s wi h he o al s a ion e e ence) is wo se by he ac o 2-3, bu signi ican ly be e han he indus ial came a. 4. SYSTEM CALIBRATION The sys em calib a ion o he HCU 3D-IMAGER includes he ollowing s eps: (a) he adap ion o he came a in he nodal poin o he pano ama head and (b) he de e mina ion o he eccen ici y and o ien a ion o he wo senso s (came a and lase ange inde ). The en ance pupil o he came a lens mus lie accu a ely in he nodal poin , since he ho izon al and e ical o a ion o he came a mus be pe o med a ound he en ance pupil o he lens. Du ing he came a o a ion a ound he en ance pupil he lase ange inde o a es simul aneously on he pano ama head. I is assumed ha he ela i e posi ion and o a ion o he lase ange inde will no change wi h espec o he came a. The e ical and ho izon al adjus men o he came a is done sepa a ely ia alignmen on a close and a emo e poin . The came a is shi ed un il he posi ion o he en ance pupil is p ecisely ixed in he nodal poin using he alignmen pla es o he pano ama head. An imp ecise alignmen leads o pa allaxes a close ange while u ning he came a, which esul s in gaps in he pano ama image. The ela i e posi ion o he lase ange inde o he came a is de ined by a shi ec o (eccen ici y), while he ela i e adjus men o he lase ange inde o he came a is desc ibed by a di ec ion ec o . Thus he ela ion be ween came a and lase ange inde is ep esen ed by a coo dina e ans o ma ion wi h i e deg ees o eedom. The shi ec o is measu ed wi h an accu acy o 1-2mm, while he de e mina ion o he di ec ion ec o is pe o med wi h an i e a i e me hod. Fo he calib a ion and alida ion o he o e all sys em wo pano ama images we e aken wi h h ee pho o se ies including 18 pho os each in he 3D es ield. The lase poin s we e measu ed wi h a o al s a ion T imble S6 (measu emen 1) and/o wo T imble 5603 DR200+ (measu emen 2a and b) as e e ence. The p ecision (s anda d de ia ion) o he o al s a ions is 2” (= 0.6mgon) and/o 3” (= 1.0mgon) o an angle measu emen (ho izon al and e ical) and 2mm + 2ppm o a dis ance measu emen wi hou e lec o . Came a I IP OP  xy [pix] SD PP [mm] C/Ch X [mm] Y [mm] Z [mm] max.  [mm] D700_20 27 857 39 0.13 0.17 6/33 0.12 0.14 0.10 0.59 D90_20 36 1089 40 0.19 0.19 6/34 0.17 0.16 0.23 0.71 IDS XS 14 365 36 0.57 0.52 6/30 0.53 1.30 0.59 5.16 RPI 1 15 460 39 0.20 0.11 6/33 0.18 0.45 0.21 1.87 I….images, IP….image poin s, OP….objec poin s,  xy ….s anda d de ia ion o image coo dina es, SD PP….s anda d de ia ion o con ol p oin s , C/Ch….con ol / check p oin s ,  XYZ….em p i ical accu ac y Table 3. S a is ics o he bundle block adjus men and empi ical accu acy o di e en came as in he 3D es ield in compa ison 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, Volume XLI-B5, 2016 XXIII ISPRS Cong ess, 12–19 July 2016, P ague, Czech Republic This con ibu ion has been pee - e iewed. doi:10.5194/isp sa chi es-XLI-B5-55-2016 58 The pola coo dina es o he lase poin s we e con e ed in o Ca esian coo dina es, so ha bo h sys ems – lase poin s measu ed by o al s a ion (sys em 1) and pano ama image (sys em 2) – a e a ailable in he same coo dina e sys em. Fo accu acy e alua ion he coo dina es o he lase poin s we e ans o med dis o ion- ee by a spa ial simila i y ans o ma ion (Helme ans o ma ion, 7-pa ame e ans o ma ion) in o he pano ama coo dina e sys em (HCU 3D-IMAGER) (sys em 1  sys em 2). Du ing an op imiza ion p ocess (a e Sansò 1973) he se en ans o ma ion pa ame e s we e de e mined as ollows: a spa ial shi ec o (3 pa ame e s), a scale ac o (1 pa ame e ) and a spa ial o a ion (3 pa ame e s). The ans o ma ion pa ame e s we e de e mined using i e con ol poin s (#CP) in each da ase . Fo he i e con ol poin s he RMS alues RMS ( ), RMS ( ) and RMS ( ) we e de e mined. The emaining 37 poin s we e used as checkpoin s (#ChP). The di e ences in he checkpoin s we e de e mined as RMS alues (RMS (X), RMS (Y), RMS (Z)). The esul s a e summa ized in Table 4. The esidues o he con ol poin s a e be ween 2mm and 5mm a e he adjus men . Howe e , he empi ical accu acy o he checkpoin s is be ween 3mm o 11mm. The easons o hese highe de ia ions a e accoun ed o by he measu emen s o he lase poin and by he p ecision o he measu ing ins umen . Since each lase poin shows as a ela i ely la ge i em in he pho os and since he bo de o he lase poin is no well de ined, he a ge ing o he lase poin cen e wi h he o al s a ion is no always comple ely p ecise (see also Fig. 5). Fu he mo e, he p ecision o he o al s a ion is limi ed due o he p ecision o he ( e lec o less) dis ance measu emen s o 2mm. M Con ol Poin s RMS [mm] # ChP Check Poin s RMS [mm] X Y Z 1 3.5 2.7 4.1 37 3.9 2.9 2.9 2a 4.8 2.6 4.4 35 10.8 3.3 6.0 2b 2.3 1.4 1.8 37 6.3 2.6 6.6 M…Measu emen , ChP…Check Poin s Table 4. Empi ical accu acy o lase poin s o he HCU 3D-IMAGER using i e con ol poin s 5. 3D INDOOR DATA ACQUISITION Fo in es iga ions in o 3D indoo mapping he calib a ed imaging sys em HCU 3D-IMAGER has been used. The image acquisi ion and he lase anging we e pe o med simul aneously, so ha he ed lase poin is isible in each single image (Fig. 5). The posi ioning o he sys em akes place in he cen e o small ooms wi h su icien iew o all ele an co ne s o as an op imal posi ioning con igu a ion, in o de o ealize op imal sigh dis ances sho e han 15m. A e le elling he sys em, i is eady o use. Fo he con olling o he sys em a lap op has been used, o ope a e he con ol compu e Raspbe y Pi ia WLAN. Figu e 6. Fi e image ows wi h 18 single images each o he gene a ion o one pano ama image Figu e 5. Pa o a single pho o (le ) wi h snapped lase poin in he magni ied sec ions (cen e and igh ) 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, Volume XLI-B5, 2016 XXIII ISPRS Cong ess, 12–19 July 2016, P ague, Czech Republic This con ibu ion has been pee - e iewed. doi:10.5194/isp sa chi es-XLI-B5-55-2016 59 One single da a se consis s o a lase dis ance and a pho o (Fig. 5, igh ), which is sa ed on he Raspbe y Pi compu e oge he wi h he o ien a ion pa ame e s o he aken pho o (ho izon al angle in deg ees om ixed as e and e ical angles by eading he scale). Depending upon desi ed imaging con igu a ion (image o e lap) all ho izon al pho og aphs a e accomplished wi h one e ical angula se ing. Fo a s anda d oom 18 ho izon al pho og aphs a e cu en ly used, which co esponds o an o e lap ange o 51%. These 18 pho og aphs a e acqui ed a di e en inclina ions, e.g. wi h he 5 inclina ion s ages +60°, +30°, 0°, -30° and -50°, in o de o be able o compu e a ull- sphe ical pano ama (Fig. 6). A lowe inclina ion is no necessa y, since hese pho os will mainly show he pano ama head as well as he ipod. A e comple ion o he pho og aphy 360° sphe ical pano ama images wi h a leas 12000 pixels o ho izon al esolu ion we e gene a ed om he indi idually-acqui ed images. The ho izon al and e ical angle, which a e de ined in a p ojec ile, we e used as app oxima e alues o he de e mina ion o he exac o a ions. These adjus ed o a ion angles a e he basis o he calcula ion o he coo dina es o each lase poin in objec space. The p oduc ion o a pano ama p oceeds in h ee s eps: 1. De e mina ion o gene al ie and special ie poin s on e ical lines: Since he adjus men o he indi idual images canno ake place wi h su icien accu acy, he pic u es a e linked by homologous poin s in neighbou ing images. 2. Op imiza ion o he o ien a ion pa ame e s: Using hese ie poin s he de e mina ion o image o ien a ion is in a subsequen op imiza ion p ocess. 3. S i ching o he single images: Using he imp o ed image o ien a ion he indi idual pho os a e s i ched o a pano ama including b igh ness balancing. Fo he s eps 1 and 2 he pano ama ools (h p://pano ools.sou ce o ge.ne /) and o s ep 3 he CAU S i che ( om he Ins i u e o Compu e Science o he Ch is ian-Alb ech s Uni e si y Kiel) ha e been used. Addi ionally, he CAU S i che compu es Ca esian 3D coo dina es o each lase poin . Fo he ollowing 3D modelling all cha ac e is ic pa ame e s a e w i en in o a p ojec ile and a e he e o e consis en ly documen ed. 6. 3D MODELLING Using only one indi idual pano ama image a scaled econs uc ion o an indoo objec ( oom) is no possible. Fu he mo e he 3D coo dina es o he lase poin s (see Fig. 7 small blue image), which ep esen he 3D oom as a e y spa se poin cloud, do no allow a comple e econs uc ion o he objec . In addi ion no all lase poin s a e p ojec ed on ele an plana su aces. Only by combining he pano ama image and a leas one Figu e 7. Gene a ed pano ama image and spa se poin cloud o he lase poin s (small blue image) Figu e 8. 3D modelling – .l. . . de e mina ion o he g ound loo , de e mina ion o a wall, de e mina ion o he window opening and example o a simple e-cons uc ed oom 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, Volume XLI-B5, 2016 XXIII ISPRS Cong ess, 12–19 July 2016, P ague, Czech Republic This con ibu ion has been pee - e iewed. doi:10.5194/isp sa chi es-XLI-B5-55-2016 60 lase dis ance and by conside a ion o (geome ical) shape condi ions can he oom be cons uc ed. The ollowing shape condi ions suppo he modelling p ocess: De ini ion o spa ial planes ( loo , ceiling and walls) o pe pendicula i y o wall planes o loo plane o ceiling. Wi hin hese planes 2D su aces ( ec angles, iangles o ci cles) can be speci ied using he pano ama image in he backg ound. The su aces can ep esen walls, ceilings o openings (doo s, windows). The in e sec ion o g ound loo , ceiling, and/o wall planes gene a es s aigh lines di ec ly, which can be used as auxilia y lines o he cons uc ion o he oom. The lase anging has essen ially he ask o speci ying he scale o he econs uc ion. On he basis o hese abo e men ioned model assump ions he company ETU so wa e GmbH in Cologne has de eloped he 3D modelling so wa e 3D BIS o manual modelling o in e io s. Du ing he econs uc ion p ocess addi ional in o ma ion can be speci ied, e.g. he labelling o objec pa s (g ound loo , ceiling, wall, …), he ma e ial (s one, wood, PVC, …) and he s a us (OK, ake down, epai , …). The so wa e adminis e s he indi idually- acqui ed indoo da a and he associa ed me ada a, and au oma ically compu es he solid and squa e measu e. Fo da a exchange he cons uc ed 3D model can be expo ed in he COLLADA o ma . The econs uc ion o a oom wo ks wi h he 3D modelling so wa e 3D BIS as ollows: On he condi ion ha he HCU 3D- IMAGER has been se up pe pendicula ly on he g ound loo and lase measu emen s o he g ound loo ook place, he g ound le el can be cons uc ed. Wi hin he g ound plane he loo plane can be cons uc ed nex by he de ini ion o 2D su aces using he pano ama image in he backg ound (Fig. 8 igh , yellow su ace). On he condi ion ha he walls a e pe pendicula o he g ound loo plane, he wall planes can be cons uc ed a e wa ds (Fig. 8 cen e le , yellow su ace). Wi hin he wall planes wall su aces and openings (windows and doo s) Figu e 9. Gene a ed pano ama including depic ed lase poin s and modelled oom wi h slan ed walls Figu e 10. Gene a ed pano ama and modelled oom which is unde cons uc ion 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, Volume XLI-B5, 2016 XXIII ISPRS Cong ess, 12–19 July 2016, P ague, Czech Republic This con ibu ion has been pee - e iewed. doi:10.5194/isp sa chi es-XLI-B5-55-2016 61 can be de e mined (Fig. 8 cen e igh , yellow su ace wi h opening). A e all walls a e cons uc ed, he modelling o he oom can be comple ed wi h he cons uc ion o he ceiling. The inal 3D model can be expo ed o be u he p ocessed o isualized in ano he 3D modelling so wa e (e.g. Ske chUp) (Fig. 8 igh ). Fig. 9 shows he gene a ed pano ama pho o including he isible lase poin s o ano he es objec and he cons uc ed oom, while Fig. 10 ep esen s he gene a ed pano ama image and he cons uc ed oom, which is in p ocess o cons uc ion. 7. CONCLUSION AND OUTLOOK The Labo a o y o Pho og amme y & Lase Scanning o he HCU Hambu g de eloped a low-cos 3D imaging sys em o in e io s, which does no cos mo e han 1000 EUR using he cu en sys em s uc u e. The da a acquisi ion o in e io s using he HCU 3D-IMAGER, which is equipped wi h a Raspbe y Pi came a and a Bosch lase ange inde on a pano ama head, cu en ly akes app ox. 12 minu es o a manual se ies o indi idual pho os a ound 360° and lase dis ance measu emen s o each image. The Raspbe y Pi came a used has small dis o ion alues up o 20m, bu an uns able p inciple poin (changes up o 50m). Ne e heless, he accu acy po en ial o he Raspbe y Pi came a ob ained a e simul aneous came a calib a ion does no qui e each he pe o mance o a digi al SLR came a. A e sys em calib a ion an accu acy o app ox. 1cm o single poin s can be expec ed o a s anda d in e io wi h maximum dis ances o 10m. 20 ooms ha e al eady been eco ded using his sys em and all could be econs uc ed om he pano ama images and dis ance measu emen s using he so wa e 3D-BIS. In he u u e he 3D indoo mapping will wo k wi h a mo o ized and compu e -con olled pano ama head, aking ewe pho os in wo minu es pe s a ion. I is planned ha he wo k low om pano ama p oduc ion o objec cons uc ion will be ully suppo ed by au oma ic image p ocessing ope a ions. The assump ion he e is ha he eco ded oom is mos ly o med om simple wo dimensional elemen s, and ha he p edomina ing pho o cha ac e is ics consis o a la ge ex en o ho izon al and e ical s uc u es (Manha an Wo ld). ACKNOWLEDGEMENTS The suppo o he p ojec by ZIM ( he cen al inno a ion p og amme o small and medium-sized en e p ises) – a p og amme o he Fede al Minis y o Economic A ai s and Ene gy – is g a e ul acknowledged. REFERENCES Clauß, U., 2011. Die lase pho og amme ische Objek au nahme – Ein e izien e Ve ah ensansa z ü iele Au mess- und Modellie au gaben. Pho og amme ie – Lase scanning – Op ische 3D-Mess echnik, Bei äge de Oldenbu ge 3D-Tage 2011, Luhmann, T. & Mülle , C. (Eds.), Wichmann - VDE Ve lag, pp. 205-212. Clauß, S., 2012. Ra ionelles 2D-Au maß mi RODEON me ic. Pho og amme ie – Lase scanning – Op ische 3D-Mess echnik, Bei äge de Oldenbu ge 3D-Tage 2012, Luhmann, T. & Mülle , C. (Eds.), Wichmann - VDE Ve lag, pp. 255-262. Fangi, G., 2007. The Mul i-image sphe ical Pano amas as a ool o A chi ec u al Su ey. XXI In e na ional CIPA Symposium, 01-06 Oc obe 2007, A hens, G eece. Hen y, P., K ainin, M., He bs , E., Ren, X. & Fox, D., 2014. RGB-D mapping: Using dep h came as o dense 3D modeling o indoo en i onmen s. Expe imen al Robo ics, Sp inge Be lin Heidelbe g, pp. 477-491. He ing, H., 2012. Un e suchungen zu Ge ä e- und Modell- genauigkei des lase pho og amme ischen Au nahmesys ems RODEONme ic. Pho og amme ie – Lase scanning – Op ische 3D-Mess echnik, Bei äge de Oldenbu ge 3D-Tage 2012, Luhmann, T. & Mülle , C. (Eds.), Wichmann - VDE Ve lag, pp. 286-293. Kalusche, W., 2011. G und lächen und Planungskennwe e on Wohngebäuden. Inno a ion im Baube ieb, G alla & Sunde meie (H sg.), We ne Ve lag Köln. Ke s en, T., Ace edo Pa do, C. & Linds aed , M., 2004. 3D Acquisi ion, Modelling and Visualiza ion o no h Ge man Cas les by Digi al A chi ec u al Pho og amme y. The In e na ional A chi es o Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sciences, Vol. XXXV, Commission V, Pa B2, pp. 126-132. Ke s en, T., S e nbe g, H. & S ieme , E., 2005. Fi s Expe iences wi h Te es ial Lase Scanning o Indoo Cul u al He i age Applica ions using Two Di e en Scanning Sys ems. The In e na ional A chi es o Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sciences, Vol. XXXVI, PART 5/W8. Ke s en, T. & Linds aed , M., 2012. Vi ual A chi ec u al 3D Model o he Impe ial Ca hed al (Kaise dom) o Königslu e , Ge many h ough Te es ial Lase Scanning. Eu oMed 2012 - In . Con e ence on Cul u al He i age, Ioannides, M.; F i sch, D.; Leissne , J.; Da ies, R.; Remondino, F.; Ca o, R. (Eds.), Lec u e No es in Compu e Science (LNCS), Volume 7616, Sp inge - Ve lag Be lin Heidelbe g, pp. 201-210. K ispel, U. E e s, H. L., Tamke, M., Viehause , R. & Fellne , D.W., 2015. Au oma ic Tex u e and O hopho o Gene a ion om Regis e ed Pano amic Views. In e na ional A chi es o Pho og amme y, Remo e Sensing and Spa ial In o ma ion Sci- ences, Volume XL-5/W4, Gonzalez-Aguile a, D., Remondino, F., Boehm, J., Ke s en, T., Fuse, T., (Eds.), 6 h In e na ional Wo kshop 3D-ARCH 2015: "3D Vi ual Recons uc ion and Visualiza ion o Complex A chi ec u es", 25-27 Feb ua y 2015, A ila, Spain, pp. 131-137. P esseBox, 2005. 90 P ozen de Immobilien haben alsche Flächenangaben. P essemi eilung BoxID 46989, h p://www.p essebox.de/inak i /eu oma com-public- ela ions- gmbh/90-P ozen -de -Immobilien-haben- alsche- Flaechenangaben/boxid/46989 (access 21 Ma ch 2016). Sansò, F., 1973. An exac solu ion o he o o- ansla ion p oblem. Pho og amme ia 29, pp. 203-216. Tamke, M., Blumel, I., Ochmann, S., Vock, R. & Wessel, R., 2014. F om poin clouds o de ini ions o a chi ec u al space - po en ials o au oma ed ex ac ion o seman ic in o ma ion om poin clouds o he building p o ession. P oceedings o he 32nd eCAADe Con e ence, E. M. Thompson (ed.), Vol. 2, pp. 557-566. Wagne , D., 2015. Falsche Flächenangaben on Immobilien be gen Po en ial s a Risiken ü Bes andshal e , In es o en, Eigen üme , Ve wal e und In e essen en. h p://www.kessle solu ions.de/ ileadmin/ edak eu e/CAD/Fach a ikel_01_Immobilien onds.pd (access 08 Ap il 2016). Zhu, Z. & Donia, S., 2013. Po en ials o RGB-D Came as in As- Buil Indoo En i onmen Modeling. ASCE In e na ional Wo kshop on Compu ing in Ci il Enginee ing, pp. 23-25. 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, Volume XLI-B5, 2016 XXIII ISPRS Cong ess, 12–19 July 2016, P ague, Czech Republic This con ibu ion has been pee - e iewed. doi:10.5194/isp sa chi es-XLI-B5-55-2016 62