Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Modeling and calibration of pulse-modulation based ToF imaging systems

: Süss, Andreas; Varga, Gabor; Marx, Michael; Fürst, Peter; Gläsener, Stefan; Tiedke, Wolfram; Jung, Melanie; Spickermann, Andreas; Hosticka, Bedrich J.


He, Sailing (Hrsg.) ; Society of Photo-Optical Instrumentation Engineers -SPIE-, Bellingham/Wash.:
Smart photonic and optoelectronic integrated circuits XVIII : 16 - 18 February 2016, San Francisco, California, United States
Bellingham, WA: SPIE, 2016 (Proceedings of SPIE 9751)
Paper 975117, 9 S.
Conference "Smart Photonic and Optoelectronic Integrated Circuits" <18, 2016, San Francisco/Calif.>
Fraunhofer IMS ()
pulsed modulated Time-of-Flight; imaging; PM-iToF; inverse problem; regression; cross-validation

Conversely to the continuous wave indirect time-of-flight (CW-iToF) imaging scheme, pulsed modulation ToF (PM-iToF) imaging is a promising depth measurement technique for operation at high ambient illumination. It is known that non-linearity and finite charge-transfer speed impact trueness and precision of ToF systems.1–3 As pulses are no Eigenfunctions to the shutter system, this issue is especially pronounced in pulsed modulation.2, 3 Despite these effects, it is possible to find analytical expressions founded on physical observations that map scenery parameters such as depth information, reflectance and ambient light level to sensor output.3, 4 In the application, the inverse of this map has to be evaluated. In PM-iToF, an inverse function cannot be yielded in a direct manner, as models proposed in the literature were transcendental.3, 4 For a limited range an approximating linearization can be performed to yield depth information.5 To extend the usable range, recently, an alternative approach that indirectly approximates the inverse function was presented.6 This method was founded on 1D doping concentration profiles, which, however, are typically not made available to end users. Also, limitations of the 1D approximation as well as stability are yet to be explored. This work presents a calibration methodology that copes with detector insufficiencies such as finite charge transfer speed. Contrarily to the state of the art, no prior knowledge on details of the underlying devices is required. The work covers measurement setup, a benchmark of various calibration schemes and deals with issues such as overfitting or defect pixels.