Document Server@UHasselt >
Research publications >
Please use this identifier to cite or link to this item:
|Title: ||Mobile Structured Light: Reconstruction using a Signal Processing Approach|
|Authors: ||Hermans, Chris|
|Advisors: ||Bekaert, Philippe|
Van Reeth, Frank
|Issue Date: ||2011|
|Abstract: ||In this dissertation we introduce a new set of 3D shape acquisition methods,
which we have called mobile structured light. Unlike the more classical structured
light methods, in which a static projector illuminates a (static) scene with a
variety of time-varying illumination patterns, our proposed technique makes use
of a mobile projector emitting a single static sinusoidal illumination pattern.
This projector, which we refer to as a sliding projector, is translated at a constant
velocity in the direction of the projector’s horizontal axis. Optionally, we use a
secondary illumination pattern, a De Bruin pattern tailored onto the original wave
pattern, to allow for online extrinsic calibration of our setup.
Illuminating the object in this manner allows us to perform fast per pixel
computations, in which we can analyze the incoming illumination using traditional
methods from signal processing literature. By employing Fourier analysis to
decompose the observed illumination sequence into a corresponding set of frequency
components, we are able to recover pixel depth. We have observed that there exist a
linear relationship between the depth of a scene point, defined as the distance between
the projector’s principal plane and the scene point, and the frequency of the observed
illumination sequence in the corresponding pixel. Thus, either with precise precalibration
of the setup, or using a secondary pattern for extrinsic post-calibration,
we are able to convert these depth values into a valid 3D reconstruction.
As we have explicitly cast depth estimation as a signal processing problem,
we are able to borrow from the vast literature that exists on the discussed topics.
In this dissertation, we have examined the influence of applying frequency refinement
techniques from the single tone estimation problem to our own approach, and
noticed a significant improvement in accuracy and reduction in required frames.
Furthermore, we have shown compatibility between new developments in the area
of compressed sensing, which could potentially be applied to a pre-calibrated
implementation of our technique.The proposed approach has several advantages over classical structured light
methods. As the method performs depth estimation on a per pixel basis, it is able
to preserve sharp edges in the produced depth image. Furthermore, unlike classical
structured light methods, the quality of our results is not limited by projector or
camera resolution, but is solely dependent on the temporal sampling density of the
captured image sequence. Additional benefits include a significant robustness against
common problems encountered with structured light methods, such as occlusions,
specular reflections, subsurface scattering, interreflections, and to a certain extent
Finally, we have discussed the dual relationship between epipolar plane image
analysis, a passive shape acquisition technique which uses a linearly translating
camera in order to capture the scene structure, and the proposed mobile structured
light method. EPI analysis exploits the large amount of inherent structure in the
recorded image volume to recover scene structure, relying on data redundancy and
a color consensus matching criterium to establish cross-frame correspondences.
It is this same data redundancy that determines the accuracy of the observed
frequency in our proposed dual technique, without the requirements of an order
constraint or the restrictive nature of a color consistency assumption. Thus, mobile
structured light allows for the acquisition of a much larger variety of materials,
whereas EPI analysis was restricted mainly to Lambertertian surfaces.|
|Type: ||Theses and Dissertations|
|Appears in Collections: ||PhD theses|
Files in This Item:
|PhD Chris Hermans||46.04 MB||Adobe PDF|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.