Analysis of the measurement accuracy of a thermal 3D sensor for transparent objects

Transparent, translucent, jet-black, and highly reflective objects pose major challenges for 3D sensing based on visible or near-infrared pattern projection. They either need laborious surface preparation, e.g. spraying, or cannot be digitized at all. The “shape from heating” approach overcomes these limitations by projecting patterns in the thermal infrared. The absorbed radiation creates a heat pattern that is re-emitted and captured by two mid-wave infrared cameras in stereo arrangement. This paper reports on the experimental realization of a thermal 3D sensor employing sequential fringe projection. The system is characterized for accuracy and acquisition speed according VDI/VDE guideline 2634 part 2, and a dedicated test artifact is introduced to benchmark the thermal sensor against conventional 3D metrology solutions. In addition, we investigate various scanning strategies and analyze how they affect the resulting 3D data quality. • Sequential thermal fringe projection delivers 3D scans of transparent, jet-black, or highly glossy objects in few seconds and without any surface preparation. • A comparison with micro computed tomography and conventional 3D sensors operating in the visible spectrum shows deviations of less than 10 μ m, while the thermal sensor is faster and requires no spray coating. • VDI/VDE 2634 sensor characterization shows high 3D measurement quality across the entire measurement volume of 160 × 135 × 100 mm 3 .