Assessment of Time-of-Flight Cameras for Short Camera-Object Distances

In this paper we have compared Time-of-Flight cameras of different vendors at object-camera distances of 500 mm, 1500 mm and 2500 mm. The aim was to find the highest possible precision at the distance of 500 mm, to estimate the change of the accuracy depending on scene-reflectivity and working distance and to investigate the possibility to use the cameras as per-pixel sub-centimeter accurate measuring devices. To this end, we have evaluated the variation of the distance measurement noise over several distances as well as the minimum noise we could achieve with each camera. As the amplitude-dependent distance error may become significantly large, we also tried to quantify it in order to estimate if it can be reduced to fulfill given accuracy requirements. We compared a Camcube3 from PMD Technologies, a Swissranger4000 from MESA Imaging and a C70E from Fotonic. All cameras showed different behaviors in terms of temporal noise, variation of noise and amplitude dependency. The Camcube showed the strongest amplitude dependent effects. The minimum standard deviations at 500 mm distance were at 4.8 mm for the Camcube, 1.6 mm for the Swissranger and 0.9 mm for the C70E.