Simulation of Time-of-Flight Sensors

Time-of-Flight distance sensors generate images in which each pixel records the distance between the camera center and the object that is visible to that pixel.

This 3D measurement information is useful for all kinds of applications, including 3D object reconstruction, obstacle detection, and driver-assist systems.

The simulation of such sensors is useful for developing new sensor variants and for producing Ground Truth information to train and test algorithms that use the 3D measurements.

From left to right: distance image and intensity image of a simulated ToF sensor, and the 3D point cloud derived from the distance image:
Simulated ToF distance image Simulated ToF intensity image Simulated ToF 3D point cloud

Early simulation approaches took only direct illumination into account, though some were later extended to include a rough approximation of one additional light path bounce. Simulators based on path tracing solved that problem by simulating full light transport, but since today's path tracers cannot keep precise distance information per path, approximations such as distance histograms or binning were employed.

WurbltPT instead simulates full light transport and has the capability to record precise distance information per light path. See Rendering Light Propagation for details. This makes previous approximations obsolete.

A sensor model can be fed with this precise light propagation information to produce realistic simulation results. The C++ class interface that needs to be implemented is very simple: see sensor.hpp. It allows different classes of sensors to be simulated, for example Amplitude-Modulated Continuous-Wave (AMCW) ToF sensors and Pulse-Based ToF sensors.

See sensor_tof_amcw.hpp for the implementation of the AMCW ToF sensor model described in the following paper:
M. Lambers, S. Hoberg, and A. Kolb. “Simulation of Time-of-Flight Sensors for Evaluation of Chip Layout Variants”. In: IEEE Sensors Journal 15.7 (July 2015), pp. 4019–4026. doi: 10.1109/JSEN.2015.2409816 (PDF, DOI, Code, Bib).
The wurblpt-tof-example uses this model in a dynamic scene to simulate the typical motion artefacts of AMCW ToF sensors.