Drawing Accurate Ground Plans Using Optical Triangulation Data Kevin Cain INSIGHT kevin@insightdigital.org Abstract Here we consider optical triangulation scanning as a means of creating permanent architectural archives in the form of accurate ground plans and other orthographic views We present plan drawings created with laser scan data and use these documents to make comparisons with existing documents Finally, we present a new technique for decreasing the laser scanning field time required to create plans and other views Preface This brief paper introduces an effort to create accurate ground plans for a Fatimid-era monument in Cairo The Egyptian Antiquities Project of the American Research Center in Egypt (EAP, ARCE), having undertaken conservation of the Zawiya and Sabil of Sultan Farag ibn Barquq (Al-Dehisha) in Cairo, Egypt, requested digital documentation as a basis for their physical restoration This monument, shown below, is listed as Supreme Council of Antiquities Islamic and Coptic Antiquities Monument No 203 Philippe Martinez École normale supérieure / INSIGHT pmartine@ens.fr perspective in the final image Also, the lens introduces distortions that make it difficult to extract accurate drawings or measurements The orthographic drawings commonly used by architects and engineers are drawn without perspective in order that the relationships between any given points on the drawing can be measured at a constant scale During optical triangulation scanning, a sensor measures the distance from the scanner to a specific point on a given object By making these measurements, the relationship between the gathered points can be expressed in the constructed 3D space of the computer By taking a many measurements, a “cloud of points” emerges that accurately describes the subject being scanned Finally, when rendered from the point of view of a synthetic camera in the 3D world space of the computer, digital drawings are generated Our Work at Farag ibn Barquq The scanning and registration process at Barquq involved multiple viewpoints, which were later correlated Fig 1—Digital photograph of north faỗade, Farag Ibn Barquq A Simple Review of Optical Triangulation Scanning Practice Laser scanning can be used to create a representation of an object in space, but the construction of this image comes by a very different method than is used in conventional photography When a conventional photo is taken, the image is captured through a lens The specifications of the lens determine the observed 1063-6919/03 $17.00 © 2003 IEEE Fig 2—Digital photograph of the monument’s west faỗade Four main techniques were used to document selected parts of the mosque: photography, acetate drawings, acoustic measurement, and laser scanning Conventional ink drawings on acetate were used to record the epigraphy set in limestone near the mosque’s entry Freehand drawings were made of all areas in the interior and notated with measurements as they were taken A handheld digital measure was used to verify the accuracy of the laser scan data The device produced results accurate to (+/- cm); these objective point-topoint measurements were then compared to the measurements drawn from scanned data 3.1 Scanning Methodology the narrow passage to Area C At the same time, the scanner itself had to be positioned so that it could see the spheres in Area A and C, which further limited the placement of spheres Finally, it was crucial that all spheres be left in place during scans, and that all spheres were named in the computer for registration purposes Viewpoints and their corresponding objects were planned on paper before scanning The viewpoints needed for comprehensive documentation of the monument as a whole were first worked out on paper The main interior viewpoints (112) and exterior viewpoints (A-G) are shown below (Fig 3), in plan view Fig 4—Scanning viewpoint #1, with a red reference sphere In Fig 4, it is possible to see one red reference sphere on a tripod, and two spheres placed on the floor All viewpoints and their corresponding reference spheres were planned on paper before scanning Fig 3—Primary scan viewpoints for the interior and exterior The goal of creating an integrated interior and exterior model presented unique problems because of the number of common reference objects required to survey the entire monument For the method we used, three distinct entities (spheres or otherwise) are needed to accurately reconcile two viewpoints Care was used to place reference objects where they could be seen in as many viewpoints as possible In the case of scan #3, for example, the spheres in Area A had to be visible through Fig 5—View from the roof of Barquq towards the scanner and team below; note reference sphere on ledge The requested final output for the project was orthographic drawings of the monument (plans, centerline sections, front elevations, and details) as well as digital reconstructions of the mosque at different phases of its history Accordingly, different methodologies were used for the interior and exterior scans For scans #1-12, the goal was to capture an accurate cross-section of all interior walls; these scans were used to construct the plan view and did not require the highest resolution Viewpoints A-G, the exterior scans, required scanning over larger distances For the north and east facades, it was also important to capture the entire surface ARCE-identified areas, both interior and exterior, were scanned in greater detail as a reference for future study of the monument The specific rationale for these decisions is addressed in Section Considering the scans of Farag Ibn Barquq as a study case, the following generalizations can be abstracted: Laser scanning can be an accurate technique for full 3D documentation While it is still a challenge to deal with the large amount of data generated from 3D scans, the scanning process remains the best digital method to quickly gain comprehensive 3D documentation of a site This is especially crucial for imperiled sites, where details risk being lost before they can be documented In such “crisis” cases, 3D scan data could potentially be the only resource available to future researchers An integrated approach is helpful, balancing laser scanning with traditional techniques While 3D scanning can be accurate and rapid when compared to traditional techniques, it is important to test the data provided via traditional techniques, where possible There are also clear advantages to traditional techniques in terms of cost and time, depending of the project Technique Advantages Limitations of 3D Scanning for Cultural Heritage Projects During fieldwork, our team has found that 3D scanning hardware is inevitably delicate While scanners differ in the robustness of their performance, all require special handling and careful operation Also, the costs required to complete large-scale scans and the heaviness of the resulting data files are currently a significant problem Also, while laser scanning can achieve submillimeter accuracy, it is difficult to accurately capture epigraphy with this technique Large-scale scanners are not designed to record detailed inscriptions while smallscale scanners are not equipped to deal with the scale of a building Since epigraphy must be scanned at high resolution to capture crucial detail, heavy files are again an issue 3.2 Interior Data As shown in Fig 3, 12 viewpoints were taken of the interior walls and floor regions for the purpose of generating an accurate ground plan While it would be possible to proceed with fewer points, sampling a significant section of every wall reveals implicit angles in these vertical surfaces Once the 12 viewpoints were registered using 20 distinct entities (including registration spheres as shown in Section III), a 3D model for the whole interior plan was generated Below, two orthographic views of this point cloud model are shown: a ¾ view of the interior from above, and a traditional plan view of the point clouds Limitations Photography Low cost, Fast Limited accuracy Traditional Illustration Relatively low cost Slow, Limited accuracy QTVR / digital panoramas Allows user to navigate a 3D scene Fixed viewpoints Traditional surveying Accurate, Established technology Limited number of 3D points 3D Laser scanning Accurate, High number of 3D points High cost, Technically demanding Figure Comparison of 2D / 3D documentation types Advantages of 3D Scanning for Cultural Heritage Projects 3D scanning enables a site to be accurately measured in a relatively short amount of time 3D scanning remains the only viable way of documenting the precise measurements of a complex subject such as the Sabil’s deteriorated stalactite ceiling Importantly, scan models can be transferred to popular formats (i.e., AutoCAD) for use by architects and engineers These same files can be used as the basis for reconstructions, physical models, or object movies Provided that the files are continually migrated, 3D scanning is a permanent, durable record of the site Fig 7—An orthographic view of the interior walls as seen from above Fig 9—Detail of Area A Entry Fig 8—An orthographic plan view of the interior point clouds As described previously, the 12 interior scan viewpoints were merged into an integrated model by using reference spheres common to two or more views The proper correlation of these separate views is crucial, since the registration process can introduce errors The integrity of the data taken from each viewpoint is initially secure, resulting in a high level of confidence in the measurements made from the data When two viewpoints are registered, however, the accuracy of the union is limited by the three or more common reference entities designated To improve the accuracy of these unions, entities were created to supplement the reference spheres For the plan shown at left, the average possible error was computed as - 0.6mm The resulting ground plan, shown in Fig 11-13, also incorporates all exterior scans Fig 10—Orthographic ¾ view of Area A 3.3 Correlation with ARCE / SCA Working Drawings, Undated The following drawing compares the SCA Architectural Working Drawings for Barquq (black, dashed line) to the 2000 ground plan (shaded) Fig 11—SCA working drawings v 2000 ground plan 3.4 Correlation with Saleh Mostafa Ground Plan, 1972 The following drawing compares the Dr Mostafa ground plan for Barquq (black, dashed line) to the 2000 ground plan (shaded) Fig 12—Dr Mostafa ground plan v 2000 ground plan 3.5 Correlated Ground Reconstruction, 1917 Plan Before The following drawing compares a ground plan from 1917 (black, dashed line) to the 2000 ground plan (shaded) Note: The 1917 plan shows few exterior lines The exterior dimensions of the plans vary, and are being examined against the 2000 plan and each other On the southern faỗade, the 1974 plan understates the thickness of the southeast exterior wall, while the ARCE plan overstates this dimension The same is true of the eastern faỗade, where the southeast corner is likewise distorted Unless otherwise stated, the following comparisons are made relative to the 2000 ground plan A reference is given for each note on Fig 14 below Fig 13—1917 drawings v 2000 ground plan—note the substantial reworking of Area D 3.6 Fig 14—Key to specific notes Ground Plan Comparisons The following general notes were culled from the 2000 ground plan: For the range sampled, interior walls were observed to be vertical within apprx 5-10mm variation The interior wall of room C leans east Since the building was moved during 1922-23, there are obviously many changes between the 1917 plans and all others; it is therefore left out of direct comparisons of dimension below ENTRY (NORTHERN) FAÇADE The 1972 plan simplifies the entry contours on the entrance niche, omitting the vertical grooves at the perimeter of the entrance faỗade The angle and size of the limestone threshold and entry steps is common to all four plans; for placement, the ARCE plan matches the 2000 plan most closely The extreme right edge of the northern faỗade on the 2000 plan is drawn at less than a 90-degree angle This area is not drawn in the 1917 plan; the 1972 plan draws this angle at apprx 90 degrees and the ARCE plan show the angle at greater than 90 degrees 13 AREA A 14 The ARCE plan closely matches the 2000 plan dimensions; when rotated clockwise to correct for angle, the 1972 plan also fits the 2000 plan The 1972 and ARCE plans shows a second header in the western passage to area B 15 AREA G AREA B The walls of the eastern passage between areas A and B are shown as flared in the 2000 plan This is a possible artifact of the passageway doors The ARCE plan differs from the other plans in showing no pitch to the south wall niche The 2000 plan agrees with the pitch drawn in the 1917 plan; the 1974 version shows the pitch running in the opposite direction AREA C & AREA D The ARCE plan omits the beveled corner present in the northern wall of area D On this point, the 2000 and 1972 plans closely match if rotated for alignment The south walls of areas D and E establish a parallel line The ARCE plan presents the south walls of areas D and G as parallel; the 1917, 1972, and 2000 plans show otherwise AREA E 10 is east of the same feature in the 1972 and ARCE plans If verified, the implication is that the rihab roof members are not perfectly parallel The bookcase in Iwan was closed during scans, accounting for lack of depth information at this point The 2000 plan indicates that the west wall is not perfectly true over its course All other plans show this wall as true The angled walls of the south window niche are more inclined in the 2000 plan than the others 16 On the ARCE plan, the north wall of Area G is not parallel with the south niche wall of Area B The other plans agree on this point On the 2000 plan, the line of Area G’s north wall is also shown as parallel with the north wall of Area E Section Views The construction of sectional views follows the methodology described for generating the 2000 ground plan Again, using the scan data, dimensions can be extracted as accurate, orthographic drawings 4.1 Iwan and Prayer Hall Sections The EAP team expressed interest in documenting the original beams in Area F The beams are deteriorated and irregular, as seen in Fig 15-16 The surface and volume of each beam is unique, a fact that can be appreciated as easily by rotating the digital scan model as by direct observation at the site Note fragments of the marble encrustation in the scan model, above the closed cupboard and doorway The ARCE plan properly shows the relationship between areas E and G; the 1972 plan does not AREA F 11 12 The niches along the east wall are not shown in the ARCE plan, and were not scanned for the 2000 plan They are shown on the 2000 plan according to their position on the 1972/1917 plans A boundary return in the north wall divides the room into a rihab and Mihrab section The location of this return on the 2000 plan Fig 15—Iwan point cloud Below, a scan data section is drawn at the centerline of Area F (Fig 16) A complete orthographic section drawn from this data is shown in Fig 18, including major beam measurements (As with the 2000 ground plan, dimensions were verified using manual measurements Fig 16—Iwan roof section Since the character of each beam changes over its run, a single section view will only partly describe the three-dimensional detail present However, it is possible to take sectional views at regular points along length of the beams as a way of sampling the changing dimensions of the beams As shown in Fig 17, below, the total point cloud has been divided into three smaller areas for analysis Using this approach, section views can be generated at any desired interval Fig 18—Prayer Hall Centerline Section 4.2 Sabil Section Using a similar subdivision approach, section views can be generated at any desired interval for other areas of the mosque, such as the Sabil ceiling In Fig 19, below, an orthographic centerline section view of point cloud data is shown Fig 17—A subdivided view of the Sabil section data Fig 19—Ceiling section view of Sabil point cloud data recorded The entire inscription, with translation, is shown in Fig 23 below Fig 22 Extracted line drawing with gradient to accent the restored, rightmost block Fig 20—A contrasting axonometric view of the Sabil ceiling, seen from below Epigraphic and Faỗade Documentation In addition to recording the building exterior with laser scanning, traditional line drawings were completed for principal inscriptions on the north faỗade 5.1 Texts The epigraphy on the north entry walls is treated below (In Fig 21, note the restoration on the right third of the text: the restored block is clearly revealed by the color and lighting in this photograph.) Fig 23Epigraphy and translation from north faỗade Epigraphy running along the top of the north and east walls was digitally photographed and assembled 5.2 Fig 21Epigraphy on north faỗade For comparison, a fragment of the extracted line drawing on acetate is shown below After the full-scale sheets of acetate were inked, they were digitally Faỗade Reconstruction An area of inlaid and mosaic work directly above the main entry was selected for a simple digital reconstruction This area is shown in Fig 24 high resolutions as a baseline for future study of environmental effects on the structure 6.1 Digital Architectural Archive Easily quantifiable landmarks were designated on the exterior of the mosque; these regions were then scanned and archived as 3D clouds of points Areas recorded in the way include the upper Muqarnas rows along the eastern faỗade and the main entry portal area on the north faỗade By comparing the data recorded in March 2000 with future 3D scans (of any kind) it would be possible to study changes in the structure over time Fig 24 (left)—Entry portal Below, a photo showing the existing condition is contrasted with a reconstructed image (Fig 25-26) Color reference for the reconstruction came through research, informed by study of mosaic fragments elsewhere on the exterior of the mosque Fig 27 (left)—Point cloud view of the north faỗade entry portal, (right)Photographic reference A New Technique for Increased Speed and Accuracy in Site Scanning During viewpoint framing for long-range laser scanning, nearly all current scan control software assumes a uniform bounding box selection (parametric UxV) within an XYZ world Here we suggest a new system of scanner control that does not make this assumption, but instead uses active parsing of incoming points to enable automated, “subdivided” scan viewpoint framing Fig 25—A photo of the portal region 7.1 Unique Challenges in Large-Scale Scanning Because access to archaeological sites is often limited, it can be difficult to scan large sites at relatively high resolution This was true for the three-day schedule allotted for our documentation of Farag ibn Barquq Time constraints were also the case of our laser scanning at the Ramesseum in Thebes, Egypt In this case, like Barquq, our goal was to establish detailed architectural plan, section, and elevation views However, in this case we documented a site many hundreds of meters in dimension Fig 26—A digitally reconstructed view of the same portal Exterior Documentation As requested by Dr Vincent, director of the ARCE EAP, sections of the monument exterior were scanned at Fig 29 Schematic plan view of a sample scene Fig 28 The Ramesseum, Thebes, Egypt Current survey techniques can yield apprx 1:1000 digital maps [1] However, to resolve the complex plan at the Ramesseum, we estimate a resolution of 10mm would be ideal Even through our team was able to acquire points rapidly (~1000 points/sec), it was still not practical to scan the entire site at this resolution, since this would require approximately 250 days During the initial (red) scan, our proposed method evaluates the change in δscan from point to point When the ∆δscan > Depth Threshold (a user specified value), a corner has been detected and a detailed scan is launched (shown above in green) to determine a more accurate position for this edge, which in turn allows for greater precision in view registration via ICP The Depth Threshold can be set by the user so that minor changes in distance (for instance, the uneven bricks in the wall shown at left) not trigger an automated scan refinement pass The following is pseudo code for a single scan line: 7.2 Increasing Detail via Progressive Refinement • While UVcurrent > UVend_of_line • We propose a refining technique to 1) detect edges during the scanning process, and 2) mark these border areas for rescanning at a higher resolution specified by the user This is particularly appropriate for archaeological sites such as the Ramesseum, where the large-scale structure of the site is composed of rectilinear elements (as shown above) Ours is a ‘poly-resolution’ approach that aims to concentrate detail where it is critical for establishing relevant architectural details (i.e., corners) while allowing less critical surfaces (walls) to be scanned more rapidly In the scene below, an initial synthetic scan pass is shown in red Note that the actual spacing between scanned points on the target surface (PIV, or Point Interval Value) increases with the angle (θscan) and distance (δscan) of the laser head relative to the scanned surface Since the movement of a scan head is typically monotonic, as shown in Fig 29, the target PIV for the higher resolution scan must take into account θscan That is, to maintain the target PIV, the rotation increment of the scanner must decrement as θscan increments Since scan point confidence falls with increased θscan, a natural limit to θscan can be set and the program instructed to flag low-confidence points for user review If ∆δscan > Depth Threshold ! ! ! • • • • Reset laser to UVprev Frame UVprev −> UVcurrent Scan framed region using PIVhighres ! If θ scan > UserLimit, flag these ! points as low confidence for the user end (If) Else increment UVcurrent to next scan point end (If) end (While) End Of Algorithm References [1] DURAN, Z., TOZ, G 2002 Documentation and analysis of cultural heritage by means of photogrametric methods and GIS, Istanbul Technical University, Division of Photogrammetry ... Working Drawings for Barquq (black, dashed line) to the 2000 ground plan (shaded) Fig 11—SCA working drawings v 2000 ground plan 3.4 Correlation with Saleh Mostafa Ground Plan, 1972 The following drawing. .. drawing compares the Dr Mostafa ground plan for Barquq (black, dashed line) to the 2000 ground plan (shaded) Fig 12—Dr Mostafa ground plan v 2000 ground plan 3.5 Correlated Ground Reconstruction, 1917... following drawing compares a ground plan from 1917 (black, dashed line) to the 2000 ground plan (shaded) Note: The 1917 plan shows few exterior lines The exterior dimensions of the plans vary,