surveying simple technique

SURVEYING SIMPLE TECHNIQUE 1

Antonio ÁLVARO TORDESILLAS (1) Escuela Técnica Superior de Arquitectura. Universidad de Valladolid. España [email protected]

Abstract The graphic surveying of works of art can be seen as a research method per se because their results allow us to deepen the knowledge of our heritage, which is the indispensable basis for its conservation. In this paper we want to show what and how new low-cost techniques of surveying are capable to do, in a very simple manner, without losing precision and fidelity; applied to the non previous drawn sculpture. In this way we are documenting them for the first time and creating a repository of three-dimensional works, as well as allowing an immediate and wider dissemination via Internet, all serve to safeguard them against any possible fatality. This technique experiments with the stereovision in two ways: first by the two parallel photos restitution and second, with the single photo made by a two lenses camera, which simplify even more the process. Keywords: surveying, photography, photogrammetry, photomodeler, low-cost.

1. Surveying simple technique by digital photography The following paper does not intend to repeat anything already written in many articles in refereed journals, or conference communications about graphic surveying of art works through laser technologies that on today, are well known and whose evolution has achieved in a surprising way to simplify both data collection and in its handling. The reason that moves us to make such surveys usually is related mainly to its conservation, so it is necessary to deepen their formal knowledge in a comprehensive manner. There is no need to lead how metric photogrammetry tries to get two and three-dimensional metric information from photographic images. In recent days metric photogrammetry is responsible for generating distortion and projection corrected images suitable for multiple uses, mapping or not, as multimedia, animation photorealistic, environmental studies, surveys architectural, archaeological, etc. However, in a parallel way, and almost dazzling and overwhelming, scanning systems of three-dimensional space using laser beams in a fan, have greatly simplified the process, converting metric photogrammetry in something whose basis is hardly necessary to know. The job has become a mere routine work and almost handcrafted, cleaning and debugging the millions of points (x, y, z), with its colour component (R, G, B), are located in space in an almost magical and immediate. To put the laser in the place to restore brings us more than half an hour, for instance, compared to the three days which means developing the final point cloud. Coloured cloud from a photograph taken by the laser device itself, with a resolution and quality, usually mediocre or dressing it by matching points with another taken from a more appropriate perspective. Three-

1

dimensional models of objects registered, as such sets of points without further processing, are models that operate only visually. The laser scanner is today a common tool in architecture and engineering projects that has been transferred successfully to the field of Cultural Heritage. However, even today, involves two major drawbacks: its high price and awkward in some cases, such as its elevation over the ground. There are perhaps, other problems with the laser, as has been the case in this the Research Project, which depends substantially on the surfaces to be swept, both in texture and colour. And it is in these cases when the projective geometry, the foundation of metric photogrammetry, rises from the ashes deposited by the laser in its incendiary appearance. Obviously, the time is also elapsed for her, and is no longer necessary to have a team of restitution in pairs, both instrumental and human. The developments of computers and software have replaced those and in just a few mouse clicks we can emulate, more accurately still, that laborious work. The development of photography with his leap to digital and its dizzying technological evolution and economic recession, which have made it accessible to anyone with a capacity of accurate documentation once unthinkable, has been the accomplice of the software to get put a dangerous trip to laser. The economic question, that voice has become so important in these times of crisis, it is almost indisputable front of some equipment that divides their costs by ten. The first one that comes to mind is if accuracy is acceptable. The answer is of course subjective, as is everything we do in surveying works. The separation of points provided by a laser is much less than that gives us the size of a pixel of a digital picture, for a same distance. However, either method, the accuracy is decided by the surveying author in the taking of some preliminary decisions, and not the machine. Both the laser and the photo can approach both the object as you want, so it does not seem to be being valued by the decision of why choose, when one of the variables is economic decision-making. Regarding the economic issue, a new phenomenon is emerging and developing also thanks to the rapid change in Internet networks on speed and capacity. Although different commercial software that have appeared are already economically accessible enough to buying a laser, lately there are companies and universities that make it for free. The control over precision and errors of the models are more difficult to assess, but, according to which cases, feasible in the majority. Among the first we have evaluated some programs such as PhotoModeler Scanner, of EOS System, while among the seconds we can easily find on the internet others as Autodesk 123d, ARC 3D Webservice or My3dscanner.com team. Furthermore, we find what concerning the instability of the laser against some situations where it is necessary, either raise, or put it in places of difficult access. What anyone who has handled a laser scanner will know how extremely important the stability of the instrument, for which you invest a large preparatory time. It shall be ensure that the laser does not move during the minutes that is sweeping the area. If the object to raise the elevation will require several meters over the floor of the scanner (high weight and fragility), or becomes in motion, this required stability cannot be done except for very bulky and expensive auxiliary resources, which may not always be achievable. The survey and construction of a three-dimensional point cloud by means of photography and the software that processes, solves these drawbacks of the laser in an almost immediate. Any elevation of the camera, much lighter and less fragile, can be solved, for example, through a telescopic aluminium pole, as we have already done with colleagues at the Università degli Studi di Salerno, in Italy. The stability is resolved with a high shutter speed shot of a simple photo. The number of pictures to make, in addition to being fast and independent, is unique and can be easily sweep the entire area to cover and a bonus in the different distances in order to allow up to work with different grades of detail, as mentioned, and allow to approach those holes where a laser and an operator would never more or less skilful invention should be filled in a phase of polishing. And we talk of telescopic poles, to quote the cheapest, because we can complicate as much as we want: balloons, microdones, etc. The more complex work, in a striking, is to make good pictures. However, it is so basic and so systematized that is not a major difficulty. We know that the confluence of a series of photographs, inwardly and outwardly oriented, i.e. with knowledge of the data defining the conical reference system, are able to locate in the

space as many points as desired, provided they appear in two or more pictures. So taking pictures have to adapt to these conditions of projective geometry: we know the focal length, for which it is necessary that the focus does not move in each cloud to be processed, always will point to a centre or original most always imaginary, and try to make our route, either along, or around the object to be lifted, the distance between photos (base) is related to the distance in order to restore conditions to turn in stereographic pairs, both horizontally and vertically. As a general rule we adopt such a relation (base/distance) is close to 0,15. And of course, it is essential that the images are perfectly focused so that the software can easily and certainly recognize the pixels in each pair related (Fig. 1, 2).

Fig. 1, 2: Position of the photographs on textured 3d model built with PhotoModeler Scanner of EOS System; and view of the mesh calculated by 123d of Autodesk and position of some photographs. In both, we see how the pictures surround the sculpture in two parallel rings separated in height each other the corresponding base for the relationship base/distance of 0,15.

The application objects with which we have tested this new technology have been some sculptures raised by features that were able to test it. Aspects such as size or different textures do not have meant much problem, except for those whose skin is of a uniform colour, especially black, or reflective or bright as some pieces of alabaster or glass. To be decided to work with sculptures rather than larger architectural pieces responds to various reasons: - They allow us to check as closely as possible margins of error and can compare the original measures taken by the traditional method, with results obtained by the laser scanner, or by reimbursement photo. - More and more is being applied photogrammetric restitution for the study and documentation of sculptures of artistic and historical value. But to date, it has not been applied with the same intensity on modern sculptures. - The sculptures, unlike architecture, have never had a graphical specification prior to its execution, so it lacks precise graphical information that would allow its return for research and documentation, including for damage of these parts, given the risk to many of them being in public places. His graphic documentation, through virtual models manipulated by computers, would become an insurance against any adversity, thus gaining more interest. It seems inevitable that this precaution taken ends up in all those works of recognized artistic value. In Research Project we have conducted the data collection of the sculptures with two different lasers (Faro Focus 3D and other portable ARTEC-ABACUS) and by digital photography with Canon 1000D SLR camera and Sigma 28mm fixed lens. The results are very similar, if we disregard, as we have clarified, the subjetibilidad the accuracy of the work. Similar in both simple geometry and texture pieces, as those most intricate (Fig. 3).

Fig. 3: These images show how the flat black color is a serious problem for the algorithms of calculation of any software, by not recognizing the corresponding homologous points. The images refer to PhotoModeler Scanner and 123d of Autodesk, but have also been tested and ARC 3D Webservice and My3dscanner. The laser scanner is not able to resolve a point cloud right in these cases.

Parts made of welded sheet metal and painted flat black have been impossible to build using the Faro Laser and photography, something better with the Artec. The recognition of black surfaces for the laser is, as we all know, a problem, until today, impossible to solve. Artec laser could provide these better results based on their movement and other possible reference points near, but outside the scanning surface itself. The picture is also unable to survey a cloud of points from an automatic mode given the impossibility of recognize different pixels in an area where everyone is equal. However, the metric photogrammetry PhotoModeler Scanner software is able to operate gives us the possibility to manually locate points in space, by referencing each on pictures (Fig. 4). The software is able to evaluate the error in this manual entry by two values: the Root Mean Square (RMS) of the coordinates of all points introduced and higher residual value that occurs at a point (Largest Residual pixels) with an option to correct, according to its or our discretion. The latter is close to one and the other to 0.5 pixels, represent model accuracy considerably good; consider the equivalent size of the pixel in a universal measurement unit.

Fig. 4: Model constructed manually by assigning control points in all photographs. These are both coded targets that were placed to facilitate the orientation of the cameras, as the vertices that define the geometry of the sculpture.

So only the software and photography are able to provide us with the necessary and sufficient to build the sculpture in 3d, and therefore in any of the dihedral 2d projections we want. This has proven to be done in a very economical, simple and accessible to everyone. However, we tested other possibilities such as the projection of conveniently defined mesh point’s images on flat and black surfaces, so the software recognition algorithms can find a way to associate these points in the space of an automatic way and are, therefore able to build points clouds avoiding the problem. Obviously this involves making double issue of photos, and then it would be necessary to replace them with those projected by the original texture. In a final phase of the research project, is proceeding to build a website that shows the results easily and intuitively, as if it were a virtual museum where anyone can access and walk among them. All this is being done easily from open source VRML environments, such as Blender, from Stitching Blender Foundation, or by viewing videos uploaded in YouTube or similar. We believe that the artwork’s surveying by this new photographic technique, will become, in all likelihood, a reference method; and this is one of the first trials and research made in Spain, and to works of this magnitude.

Bibliographical References [1] ÁLVARO TORDESILLAS, Antonio. Low-cost photogrammetric new tecnic by digital photography; applied to the S.A.V.E. of works by Chillida and Oteiza. In AA.VV. LE VIE DEI MERCANTI. SAVE Heritage. AversaCapri: La scuola di Pitagora editrice, 2011. ISBN: 978-88-6542-046-1 [2] BARBA, Salvatore; IANNIZZARO, Vincenzo; GIORDANO, Maria. Nuvole di punti vs nuvole di pixel. In AA.VV. LE VIE DEI MERCANTI. Rappresentarela Conoscenza. Napoli: La scuola di Pitagora editrice, 2010. ISBN:978-88-6542-014-0 [3] CUELI LÓPEZ, Jorge Tomás. Fotogrametría práctica. Tutorial Photomodeler. 1ª ed. Santander: Tantín, 2011. 355 p. ISBN 978-84-96920-93-4