High-Speed Optical 3-D Measurements for Shape Representation

Become a member or log in to view the full text of this article.

OSA Members get the full text of Optics & Photonics News, plus a variety of other member benefits.

High-Speed Optical 3-D Measurements for Shape Representation

 

figure(a) Setup for stereo-photogrammetric shape measurement with speckle projection: Two cameras, laser, acousto-optical deflector, lens, diffuser and the object in distance z. (b) Scheme of temporal correlation. Corresponding pixels share the same gray-value sequence in time (N images). (c) An accurate 3-D measurement of a circuit board in pseudo-colors, with a measurement time of 5 ms. (d) Static snapshot from the 3-D movie media showing a table tennis ball in a 20 x 20 x 20 cm3 measurement volume.

Optical 3-D shape measurements with structured illumination are old hat. They deliver accurate shape representations of nearly arbitrary objects and have become an indispensable tool in industries where multiple camera approaches inspect various kinds of objects (e.g., teeth, car doors, archaeological samples) for replication, conservation or quality control.

But facing high measurements rates—more than 1 3Dfps (3-D shape measurement per second)—many approaches lack the suitable projection device, and the common DMD-projector technique is widely used to project phase-shifted stripes at a maximum rate of 60-255 Hz. Our group thought about concepts for addressing the projection bottleneck for fast 3-D shape measurement, since cameras can offer image rates of up to 10,000 Hz. We published an idea for creating faster patterns by coherently scattering laser light at a diffuser surface.1 The concept is based on an uncommon but valuable way for solving the correspondence problem by the temporal correlation2 of statistical patterns.3

Instead of projecting exactly shifted stripe patterns in combination with gray-code images, objective speckle patterns illuminate the object under test. These patterns cover the object and are camera-independent. Different structures or speckle patterns are now generated by changing the incident laser beam direction at rates of unprecedented speed (a). This is done by an acousto-optical deflector (AOD), which diffracts 75 percent of the laser light into the first diffraction order that propagates at an angle determined by the grating period.

Because the diffraction grating period can be switched with up to 205,000Hz, different spots on the diffuser can be illuminated at this rate, as can the object with objective speckle-patterns. By synchronizing this speckle-pattern-creator with high-speed cameras, we could theoretically acquire 8,540 3Dfps using 24 images for a single 3-D reconstruction with temporal correlation (b). However, as the available cameras had a maximum frame rate of 4,700 Hz, we obtained 3-D movies with a rate of 193 3Dfps and a point localization accuracy of 35 µm, which has been determined by calculating the standard deviation of the 3-D points about a fitted reference plane. As the supplementary online media shows, moving objects can be acquired, and separate arbitrary textured objects can be reconstructed accurately. We presume this technique will enable new applications and fields such as inline inspection at assembly lines or multimedia 3-D acquisition.

In contrast to other concepts for high-speed 3-D shape measurements with structured illumination, such as binary pattern defocusing,4 this approach does not change its fundamentals or restrict the class of objects measurable; therefore, it delivers similar accuracy as slower pendants (relative accuracy, approximately 1 x 10-4). Summarizing this, the accuracy is not a function of speed and henceforth the camera acquisition rate limits the maximum 3-D shape measurement rate.


Martin Schaffer, Marcus Große, Bastian Harendt and Richard Kowarschik are with Friedrich Schiller University of Jena, Institute of Applied Optics in Jena, Germany.

References and Resources

1. M. Schaffer et al. Opt. Lett. 36(16), 3097 (2011).
2. P. Albrecht and B. Michaelis. Proceedings, 14th International Conference on Pattern Recognition 1, 845 (1998).
3. A. Wiegmann et al. Opt. Express 14(17), 7692 (2006).
4. Y. Gong and S. Zhang. Opt. Express 18(19), 19743 (2010).


Add a Comment

ADVERTISEMENT
ADVERTISEMENT