Data & Analytics

3D Digitization of a Large Model of Imperial Rome

Description
3D Digitization of a Large Model of Imperial Rome
Published
of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
    3D digitization of a large model of imperial Rome Gabriele Guidi,Laura Micoli, Michele Russo  Dept. INDACO - Politechnic of Milan, Italy g.guidi@ieee.org   Bernard Frischer Monica De Simone  IATH – University of Virginia, USA frischer@virginia.edu  Alessandro Spinetti,Luca Carosso  DET, University of  Florence, Italy spinetti@det.unifi.it   Abstract This paper describes 3D acquisition and modeling of the “Plastico di Roma antica”, a large plaster-of-Parismodel of imperial Rome (16x17 meters) created in thelast century. Its overall size demands an acquisitionapproach typical of large structures, but it is alsocharacterized by extremely tiny details, typical of small objects: houses are a few centimeters high; their doors,windows, etc. are smaller than 1 cm. The approach followed to resolve this “contradiction”, is described.The result is a huge but precise 3D model created byusing a special metrology Laser Radar. The proceduresof reorienting the large point clouds obtained after eachacquisition step (50-60 millions points) in a singlereference system by means of measuring fixed redundant reference points, are reported. Finally a proper splitting of the data set into 2x2 meters sub-areas for allowing data merging and mesh editing is shown. 1. Introduction The project discussed in this paper forms an important part of the Rome Reborn Project, an international effort tocreate a real-time digital model of ancient Rome. Thespatial limits of the Rome Reborn model will be the areaenclosed by the late-antique Aurelian Wall. For a varietyof practical reasons, work on the model commenced in1997 with modeling of the late-antique phase (ca. 400A.D.), which represents the climax of the development of the ancient city in terms of its urban fabric. The approachto modeling has been to work out from the city center inthe Roman Forum, a multi-purpose space dedicated to political, economic, religious, and entertainmentactivities.This phase of the city’s urban history is welldocumented and studied. There is even a highly-regarded plaster-of-Paris model - the so-called “Plastico di Romaantica,” housed in the Museum of Roman Civilization(Rome/EUR) - which, with the permission of theMuseum, could be used as the basis for the new digitalmodel. The model, created at a scale of 1:250, representsa three-decade collaboration of model-makers andtopographers in Rome. It was completed in the 1970s andhas not been changed since then. Figure 1. The plaster-of-Paris model of ancientRome digitized in this project For the Rome Reborn Project the advantages of usingthe Plastico are that the physical model could: (1) providean almost instant computer model of the project’s firststep (i.e. late-antique phase); (2) repurpose the Plasticoand keep it constantly updated and therefore useful tostudents and scholars in the twenty-first century; and (3)offer a total urban context for the new digital models of individual sites and monuments created by the RomeReborn Project. These new “born-digital” models - suchas the Roman Forum, Colosseum, Circus Maximus, andother key public buildings and monuments - were worthcreating despite the availability of the digital Plastico because they could be made at a scale of 1:1, could betextured photorealistically, could reflect discoveries madesince the 1970s, and could (when archaeological datasufficed) include the interior spaces as well as theexteriors. As a physical model created at a small scale andintended to be viewed from a high balcony, these were  features that the “Plastico” could not offer. The present project thus entailed creating a hybrid model of late-antique Rome that would be based on the digitizedPlastico and the new “born-digital” models of specificsites and monuments in the historic city center.The purpose of this paper is to describe the proceduresfor acquiring and generating a huge 3D model which presents several difficulties. In general three-dimensionalacquisition techniques are somewhat focused on a particular range of volumes. Most 3D scanners based onthe triangulation principle are suitable for small objectsand may generally work at distances ranging from one-half meter to few meters. Their measurement accuracyover the whole range image stays below one-tenth of amillimeter, and the uncertainty lies between 50 and 200microns [1]. On the other hand, laser scanners based onTime of Flight (TOF), used for architectural elements andlarge structures (bridges, dams, etc.), allow much larger distances to be covered (up to few kilometers). Althoughaccuracy remains high, the major drawback to TOFscanners is the loss of precision since the measurementuncertainty goes down to several millimeters. Thisabsolute value is not a problem for measurementsinvolving large structures because the relative precisionremains high, but if the structure is large and if smallfeatures must be captured, this kind of system is notusable. The “Plastico di Roma antica” lies unfortunatelyin the latter category, being a wide object (16 x 17.4meters) with houses and temples only a few centimeterstall. Therefore in this case, the use of conventionaltechniques was not feasible. The solution was found in asystem created for advanced metrology applications. Atfirst glance, the approach taken resembles TOF laser scanning, but its main improvement is in the procedureemployed for detecting the laser time-of-flight. Instead of conventional pulsed techniques, the method used for thePlastico uses a principle well known in CW radars, basedon transmission and reception of a coherent frequencymodulated wave. For this reason the system is indicatedas Laser Radar (LR). The actual 3D sensor (LR200) usedin this project is manufactured by Leica Geosystems AG,Switzerland in cooperation with Metric Vision Inc., VA,U.S.A. The use of such an advanced laser processingmethod, together with the capability of precisely re-focusing the laser beam in order to minimize its spot size,allows resolutions to be reached below 1 mm.Uncertainties of the same order can be obtained as thoseoffered by triangulation 3D scanners (from 0.1 to 0.3 mm,depending on distance), with the possibility of coveringdistances up to 24 meters. In order to minimize theacquisition time, a specific piece of software wasdesigned for managing the instrument at low level. It iscapable of focusing the laser at the beginning of eachscan-line and maintains a constant surface-to-laser distance during the acquisition.Another difficulty was data processing. Two separatesessions were planned: the first massive scan for coveringmost of the surface was performed from three acquisition points forming an equilateral triangle. The secondcampaign occurred twenty days later after the study of a“pre-model” generated after the first session. The huge point clouds obtained after each phase were reorientedinto a single reference system thanks to the measurementof fixed redundant references, and each was divided intosmaller sub-areas, 2 x 2 m each. This subdivision wasnecessary owing to the huge number of points in eachindividual scan (50-60 million). All these processes weremade with software specifically designed for this projectsince no commercial package suitable for managing sucha large number of points could be found. 2. Acquisition devices and procedures As demonstrated also in previous 3D scanning projects in the Cultural Heritage field, such as for example the 3D acquisition of Michelangelo’s Pietà [2],the “Digital Michelangelo” Project [3], or the digitizationof Donatello’s “Maddalena [4], working with 3Dscanning in a museum is often more complicated thandoing the same task in a laboratory. Additionalconstraints complicate what in normal conditions couldhave been done much more easily. Acquiring of the plaster-of-Paris model of ancient Rome was already a jobcomplicated enough, but the addition of further constraints made it almost impossible. For example, theadministration of the museum prohibited placement of any measurement machine directly over the “Plastico” inorder to eliminate the possibility that the machine, or oneof its parts or accessories, might accidentally fall onto themonument and damage it. The initial idea of using acommon laser blade scanning device mounted on a railfor covering the whole surface in parts was therefore notapplicable, and the sensor was chosen in order to satisfythis primary requirement. 2.1 Laser Radar The solution was found in a very high-quality (andhigh-cost) Laser Radar. It is capable of giving the same performance as a relatively low-cost and short-range laser  blade triangulation scanner. But the Laser Radar utilizedgives reliable results up to 24 meters from the measuredsurface. Since the only drawback of this extremely powerful system is its slow speed, a simpler triangulation- based laser sensor was also used for capturing the areasclose to the border of the “Plastico”.3D triangulation-based techniques have been directedtoward digital modeling of relatively small objects. Theacquisition of works of architectural (e.g. a cathedral, a  tower, a palace, etc.) is practically impossible using highresolution triangulation-based scanners, because of thegreat dimensions involved and of the distance of thescanner to object. Architectural monuments are usuallyacquired by laser scanners based on Time of Flight (TOF)measurement of light pulses, since they operate fromdistances from ten meters to thousands of meters, andthey can acquire millions of points in a relatively shorttime. By measuring the TOF needed by a laser pulse for going from the range camera to the surface and back again to the instrument, an evaluation of the camera-to-surface distance is performed. These data, together withthe angles determining the laser orientation, permitevaluation of the three spatial coordinates of any scanned point. Although this is a simple concept, its demands onsupport electronics are severe, since light covers about 30cm every nanosecond. Measurement of a few meters witha sub-millimeter resolution would require temporalresolution in the detection and processing of the backscattered signal better than 0.1 picoseconds.Commercial TOF systems offer range measuringuncertainty from 0.5 cm to 2.0 cm. In order for TOF laser scanners to digitize the whole surface of a structure, anumber of acquisitions taken from different points of view are needed, and their alignments requiresophisticated processing techniques [5, 6].The new 3D sensor used in this paper makes it possible to overcome all the above-mentioned limitations.It is a laser radar referred to as model LR200, which ismanufactured by Metric Vision Inc., VA, U.S.A. incooperation with Leica Geosystems AG, Switzerland. Theequipment is a TOF range camera, operating on a principle completely different from pulse propagation.Originally developed for microwave radars, the principleis known as Coherent Frequency Modulated Continuous-wave    Radar  (FM CW). The heart of the laser radar is a broadband frequency modulated infrared laser (100GHzmodulation). The upsweep and down-sweep comparison provide simultaneous range and velocity data for measurements. The single wide-aperture optical pathmaximizes signal strength and stability. Extensive signal processing extracts interference frequencies which aredirectly proportional to distance.A critical point is the focusing volume depth which inany triangulation-based system ranges from 10-20 cm for fringe projection systems based on white incoherent lightand to about one meter for laser blade systems. OftenTOF based system do not take into account the laser spotsize variations at distances very different from thefocusing range because for uses on buildings and largestructures it is supposed that ultra-high resolutions areunnecessary. Therefore in a standard TOF system thecorrelation between two adjacent measurements tend toincrease when the related laser spots are partiallysuperimposed, reducing the maximum resolutionattainable from the system (generally larger than 10 mm).In contrast, the Leica Laser Radar is built for applicationssuch as industrial metrology where the possible resolutioncan be much higher (e.g., far below 1mm), so that thelaser spot size is taken under control through a dynamicfocusing optical system. The drawback of thissophisticated focusing is represented by the time neededfor getting each measured point. In the most precisemodality, indicated as “advanced metrology”, only 2 points per second are measured.A peculiar aspect of this laser scanner is the method ituses for re-orienting the data into the same referencesystem during the acquisition stage, thus eliminating theneed of range map alignment, which is typically requiredin any modeling project. A redundant set of references,represented by steel spherical targets, actuallyimplemented with low cost “tooling balls”, are placed onthe scene and fixed in place with a custom metallic ringthat holds the ball in a specific position. The ring can beglued to some convenient spots of the scene withouttouching any delicate or old part of the work of art to bedigitized. At the first camera location the position of eachtooling ball is determined by measuring the direction of maximum laser reflectivity on the ball. Adding thedistance information and the “a priori” knowledge of the ball diameter, a very accurate estimate of the 3Dcoordinates of each reference target is obtained. For thefollowing camera locations the same targets are measuredagain to determine the rototranslation transformation withrespect to the first one.Once the new camera location is set up, each point ismeasured and automatically reoriented in the mainreference system through a procedure developed for  previous projects [4, 7], implementing the “UnitQuaternions” method [8]. This eliminated the need of atime-consuming iterative reorientation and of the relateddata redundancy (30-40% of range maps superpositionwould be usually needed). This feature significantlyspeeded up the 3D model generation. 2.2 Triangulation based laser scanner A VIVID 910 (produced by Konica Minolta, Japan)was the portable laser system used for integrating part of the 3D acquisition of the model of ancient Rome. Themaximum measuring distance is about 2 m with themiddle lens. An approximate resolution of 0.5 mm wasobtained at that distance. The corresponding measurementuncertainty, evaluated by acquiring a planar reference,was estimated on the order of 0.3 mm. This setting permitted the acquisition of range maps characterized bya suitable resolution and uncertainty with respect to theother data acquired with the laser radar. The two sets of data could be therefore conveniently merged.  2.3 Software The Laser Radar LR200 gives a simple unstructuredcloud of points in the form of an ASCII file containing along list of triplets, representing the (x,y,z) coordinates of each measured point. In order to triangulate them thesoftware package RapidForm was used (INUSTechnology Inc., Korea). It has an effective tool for creating a mesh from an unstructured cloud of pointsgenerated with a spherical symmetry, such as those produced by the LR200 scanner. For the other stepsinvolving mesh editing, both the module IMEdit of thePolyworks modeler (Innovmetric, Quebec, Canada) andthe editing module of the RapidForm package were used,depending on the kind of mesh correction to be applied.All the additional software was written at the Technologyfor Cultural Heritage Lab (University of Florence),utilizing two platforms: Matlab (The MathWorks, Inc.,USA) and Visual C++ (Microsoft Corp. USA). 3. Project planning The “Plastico” has an irregular shape and is installedin a special area, 16 x 17.4 meters, surmounted by a balcony, the floor of which is elevated about 2.7 metersrespect to the level of the city model. The internal perimeter of the balcony is covered by a balustrade 1.2 mhigh. In order to respect the limitations imposed by themuseum, the equipment had to be raised higher than the balustrade by mounting the scanning equipment over astand 1 meter high. Since the laser beam turns out to bevery inclined with respect to the main plane of the“Plastico”, the sensor-to-surface distance covered a widerange, going from 7 to 24 meters. 3.1 Scanning time estimation The critical point of measurements having large depthvariations is the Depth of Field (DOF) of themeasurement device. DOF is influenced by the laser  beam divergence that makes the spot size too large out of the focal zone, making a suitable resolution impossible toobtain. The LR200 solves this problem by dynamicallyre-focusing the laser beam in order to minimize the spotsize over the measured surface. This re-focusing isimplemented in the measurement modalities called“Metrology” and “Advanced Metrology”. With theseapproaches the operator simply has to define a perimeter over the surface to be acquired. The perimeter can include points at ranges very different each other thanks to the re-focusing of the laser beam for each new position. Theacquisition may therefore progress without any humancontrol, even if it lasts for a long time (e.g., overnight).The “price” for such a flexibility is a slow acquisition process, capable of giving, at best, only 10 points/second.With this digitizing speed, the time needed for a singlescan of the “Plastico” area would be on the order of several months (nights included). This was obviously notacceptable because of the costs involved and since itwould have entailed reducing the access of museumvisitors to the monument.In contrast, with an alternative measurementapproach, known as “Pseudo Vision”, the LR200 iscapable of acquiring hundreds of points per second. Evenif not so fast, this operating mode could be fast enough tocomplete the job in a reasonable time and was thereforeexplored as a good candidate for measuring the model of Ancient Rome. In this work modality, each measurementcan be additionally averaged over several repeatedmeasurements to lower the electronic noise responsiblefor generating measurement uncertainty. The number of measurement to be averaged is indicated by theinstrument as “stacking level”. By increasing the number of averaged points the time needed obviously increases.Preliminarily to the proper set-up definition the “speed vs.stacking level” relationship was experimentally evaluated by measuring the certified planar surface. Themeasurement uncertainty, calculated on the data set as theroot mean squared distance from the plane best fitting thenoisy 3D points, ranges for example from more than0.300 mm (no stacking) to 0.010 mm (10 averages). Thesuperior performance with respect to the “metrology”modes is obtained basically by inhibiting the real-timelaser refocusing so that the process works well only onsurfaces where variations in range are limited.Unfortunately this condition does not hold for the AncientRome model, hence a certain degree of customization of the equipment was needed. 3.2 Equipment customization The main idea for enhancing system performance wasto permit laser refocusing only at the beginning of eachscan-line, maintaining the sensor-to-surface constant for the following scan. This approach was calculated to givea scanning performance comparable to that of the“metrology” mode, allowing in the mean time overnightmeasurement to be performed, indispensable tocompleting a full scan of the monument in a reasonable period of time. In such conditions the predicted scan timewas 3-4 days for each point of view, and this wasconsidered acceptable by the museum.Unfortunately the system did not have (at least in thesystem release 3.21, used for this project), thefunctionality for performing spherical scanning, hence aspecial piece of software, capable of driving the beamalong circular trajectories, was specifically developed. Itrelies on a software library used by the systemmanufacturers for developing their measurement  software, and Metric Vision kindly provided the libraryso that we could solve our measurement problem. Thesoftware we developed was designed as a stand-alone program, capable of moving the beam along circular trajectories computed in advance, and of appending thecoordinates associated with each scan line to an ASCIIfile. Such incremental saving of data was introduced inorder to minimize any possible data loss in case of  blackouts during the long scanning sessions. In this way,a data acquisition speed of 170 points/second wasobtained, using “Pseudo-Vision” mode with stacking=2,and refocusing at the beginning of each scan-line. 4. Data acquisition Three scans were arranged for the first massiveacquisition. They were taken from three points of viewlocated approximately at the vertices of an equilateraltriangle. 4.1 Point clouds pre-processing Figure 2. The laser radar located on the balconyabove the “Plastico” According to the planning already described, theoperations made for each new position were: (1)measurement of a few fixed reference points from thenew position, made for properly reorienting all the datasets into a single coordinate system; (2) measurement of the border of the city model from the new position, in thelocal instrument coordinate system; (3) evaluation of intersections between the border of the model and circular trajectories (scan-lines at fixed focusing) spaced 2 mmapart; (4) loading of intersections in the custom scanningsoftware, interpreted as beginning and ending trajectory points; (5) scanning start.In the event, the time predicted agreed closely withthe actual measurements, totalling, with smallfluctuations, around four full days. 4.2 Point clouds subdivision The amount of data srcinating from the first 3Dscanning campaign was so heavy as to be not manageablewith current 3D software. Therefore the entire city modelwas subdivided into several sub-areas, dimensionallycompliant with the post processing packages used for the project.A preliminary estimate was approximately 50 million points (MPts) per point of view. Considering that theobject had been framed by three different points of views,about 150 MPts had to be treated at the end of the primarydata acquisition. In order to separate all the acquired datainto sub-areas, the city model has been divided into a 9 x9 grid. Figure 3. Subdivision of the area of the citymodel into manageable smaller sub-areas Since the area is 16 x 17.4 m 2 , we can consider 81 blocks, 2 m x 2 m each, covering a 18 m x 18 m area,large enough to cover all the irregular extensions of thecity model. Each block is made by 1000 x 1000 =1,000,000 points per take, or 3 MPoints once theacquisition was completed from the three points of view.Since the city model does not cover the whole fractionedarea, some blocks turn out to be empty or partiallyoccupied by 3D data. In order to have the samesubdivision for any of the three views, each cloud of  points was properly processed. Firstly the cloud wasoriented in the global coordinate system, defined as theone associated with the first scan (S 1 ). This point isnecessary in order to define the same blocks for all the
Search
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks