Scalable Multifunctional Indoor Scanning System

Tomáš Kovacovský∗

Faculty of Mathematics, Physics and InformaticsComenius University in Bratislava

Mlynská dolina, 842 48 Bratislava, Slovakiakovacovsky@fmph.uniba.sk

AbstractNowadays, we can see increased influence of computa-tional methods for capturing high dimensional naturalphenomena. One of the significant goal of computationalphotography is capturing the surface geometry. We pres-ent 3D scanning system SMISS based on a fringe patternstructured light projection for automatic 3D reconstruc-tion in metric space. We address the problem of low dy-namic range of analogous systems and offer a novel ap-proach for fast high dynamic range scanning, using sim-ple additional hardware. Our designed setup with im-plemented algorithms could be use as a flexible tool forfuture improvements of similar systems.

Categories and Subject Descriptors1.4.1 [Image Processing and Computer Vision]: Dig-itization and Image Capture—Camera Calibration, Imag-ing Geometry, Reflectance, Scanning

Keywords3D Scanning, Structured Light, HDR, Optical Calcula-tions, Camera, Projector, Computational Photography

1. IntroductionThe rising interest in digital photography in the last dec-ades results in ecosystem for sharing feelings, momentumsand a wide variety of information encoded into two dimen-sional media, the photo. But the photography itself is away of capturing a much higher dimensional world intotwo dimensional projection.

Thanks to computational performance of contemporarycomputers and sensors, we are able to take the photogra-

∗Master study programme in field of Computer Graph-ics and Geometry. Supervisor: Dr. Jan Zizka. Facultyof Mathematics, Physics and Informatics, Comenius Uni-versity in Bratislava.c© Copyright 2012. All rights reserved. Permission to make digital

or hard copies of part or all of this work for personal or classroom useis granted without fee provided that copies are not made or distributedfor profit or commercial advantage and that copies show this notice onthe first page or initial screen of a display along with the full citation.Copyrights for components of this work owned by others than ACMmust be honored. Abstracting with credit is permitted. To copy other-wise, to republish, to post on servers, to redistribute to lists, or to useany component of this work in other works requires prior specific per-mission and/or a fee. Permissions may be requested from STU Press,Vazovova 5, 811 07 Bratislava, Slovakia.Kovacovský, T. Scalable Multifunctional Indoor Scanning System. In-formation Sciences and Technologies Bulletin of the ACM Slovakia,Special Section on the ACM Student Project of the Year 2012 Compe-tition, Vol. 4, No. 4 (2012) 47-48

phy to the next level on its path to complete light trans-port capturing. This next level could be the reconstruc-tion of 3D geometry of a subject, instead of just capturinga 2D projection. The idea to capture geometry have ledto a large amount of diverse 3D scanning systems. Thesignificant group of these is based on structured light mod-ulated by a digital projector. This light is then reflectedby the scanned subject and then analysed by a digitalcamera. To understand the basic concept, we recommendwork [3]. An extensive overview of similar methods andsystems can be found at [1]. These systems has been de-veloping in a lot of different ways, dealing with differentproblems. A fundamental limitation of current systemsis insufficient dynamic range, caused by a bounded dy-namic range of the camera’s sensor. Moreover, the struc-tured light scanning systems use a point source of light,so the DR of the scene is highly affected by varying inci-dent angles of light with scanned objects. In addition, theDR of the scene is also expanded by contrast materials,commonly used due to visual appearance.

To be a part of the photography transformation and todeal with discussed limitation, we offer:

• An fully automatic and easy to use 3D scanningsystem SMISS [2], based on Gray Coded structuredlight and phase-shifting, capable of dense 3D pointcloud reconstruction.

• A novel approach to increase the dynamic range ofstructured light scanning systems, using a simpleadditional hardware.

• An co-axial optical setup, with great potential forfuture improvements of similar systems.

2. SMISSIn our work, we have built the stand-alone 3D scannerSMISS. The system works on the triangulation principleand solves the stereo correspondence problem using struc-tured light. The scanning volume is virtually divided intounique set of planes. Each of this plane receives slightlydifferent coding form the projector (from a set of struc-tured pattern images). This code is then decoded by thecamera and used to solve the correspondence problem.The coding and decoding is done automatically by thesystem. With decoded correspondence, additional infor-mation about position, orientation and optical propertiesof the camera and the projector1, we can build a densepoint cloud reconstruction with possible millions of indi-vidual measurements.

1We implement an semi-automatic calibration process forall mentioned parameters.

48 Kovacovsky, T.: Scalable Multifunctional Indoor Scanning System

Figure 1: Our co-axial setup

The final point cloud can be used for wide variety of differ-ent tasks, from solving computer vision problems (indirectuse), to pure visualisation (direct use). In our work, weanalyse both, and we built also an view dependent stereoviewer prototype for visualisation purposes.

3. HDR SMISSTo deal with the low DR of SMISS and similar systems, wecan use method for High DR image capturing using morecamera exposures. This popular method can be found at[4]. Using this technique would increase the scanning timeby a multiplication factor. In addition, the global HDRacquisition technique have limitations caused by a glare.

We offer a novel approach. Instead of increasing the DR ofthe capturing sensor, we can decrease the DR of the scene,so it will fit into camera’s DR. To accomplish this, wecould send less light to high reflective, and more light tolow reflective surfaces. We can choose camera pixel, whichcorrespond to specific surface area. This area is lit by acorresponding projector pixel. This correspondence is ingeneral equivalent to the surface reconstruction. However,if the projector’s and camera’s focal points are aligned,the camera projecotor correspondence is constant and canbe calibrated. We achieve this with an additional cameraand beamsplitter in co-axial setup, which can be found inFigure 1.

With this setup, we can capture a standard HDR imagewith the additional camera. We then analyse the imageand calculate proper scaling factors for individual regionsof a projected image (we use a few iterative relaxationsteps to deal with limitations of global HDR method).This results into weight map, which is then applied asan per-pixel scaling image for every projected pattern.Final weight map is shown in Figure 2.

The benefit of the weight map had to be applied toshrink the DR of the scene from the viewpoint of the origi-nal camera. For this purpose, we use the optical techniquefor filtering specular component of reflected light usingpolarisation, similarly to work [5]. We then capture thelambertian part of reflection with both cameras, which isindependent of viewer position.

4. ConclusionsWe have presented a solution for the problem of low DRof similar systems. This solution extends the DR of thesystem up to the product of projectors’s and camera’s DR,which exceeds 10000:1 for common available sensors. Inaddition, our solution needs just a constant increase in

Figure 2: Top left: Scene without weight mapapplied. Top right: Same scene with weight mapapplied. Bottom: The weight map

scanning time, which brings 250 % speedup for our fullpattern sequence over the trivial solution.

Our co-axial optical setup with implemented calibrationof correspondence is an easy to extend platform for futureimprovements.

In our tests of the system SMISS, we was able to reachthe resolution of 200 microns in a scanning volume of86 x 86 x 30 cm and was able to distinguish featuressmaller than 0.5 mm in depth.

Acknowledgements. We would like to thank TatraBank foundation (Nadacia Tatra banky) for financial sup-port trough grant Virtualizer: 3D Scanner for CompleteReconstruction. Greetings also belongs to ME-Inspectioncompany for hardware support.

References[1] S. S. Gorthi and P. Rastogi. Fringe Projection Techniques: Whither

we are? Optics and Lasers in EngieringOptics and Lasers inEngiering, 48(2):133–140, 2010.

[2] T. Kovacovský. Scalable Multifunctional Indoor Scanning System.Master’s thesis, Faculty of Mathematics, Physics and Informatics,Comenius University, Bratislava / Slovakia, 2012.

[3] D. Lanman and G. Taubin. Build your own 3D scanner: 3Dphotography for beginners. In ACM SIGGRAPH 2009 Courses,SIGGRAPH ’09, pages 8:1–8:94, New York, NY, USA, 2009.ACM.

[4] E. Reinhard, G. Ward, S. N. Pattanaik, P. E. Debevec, andW. Heidrich. High Dynamic Range Imaging - Acquisition, Display,and Image-Based Lighting (2. ed.). Academic Press, 2010.

[5] S. Umeyama and G. Godin. Separation of Diffuse and SpecularComponents of Surface Reflection by Use of Polarization andStatistical Analysis of Images. IEEE Trans. Pattern Anal. Mach.Intell., 26(5):639–647, May 2004.