Reconstructing 3D Objects With a Tiny Distance Sensor

There are a whole bunch of different ways to create 3D scans of objects these days. Researchers at the [UW Graphics Lab] have demonstrated how to use a small, cheap time-of-flight sensor to generate scans effectively.

The key is in how time-of-flight sensors work. They shoot out a distinct pulse of light, and then determine how long that pulse takes to bounce back. This allows them to perform a simple ranging calculation to determine how far they are from a surface or object.
However, in truth, these sensors aren’t measuring distance to a single point. They’re measuring the intensity of the received return pulse over time, called the “transient histogram”, and then processing it. If you use the full mathematical information in the histogram, rather than just the range figures, it’s possible to recreate 3D geometry as seen by the sensor, through the use of some neat mathematics and a neural network. It’s all explained in great detail in the research paper.
The technique isn’t perfect; there are some inconsistencies with what it captures and the true geometry of the objects its looking at. Still, the technique is young, and more work could refine its outputs further.
If you don’t mind getting messy, there are other neat scanning techniques out there—like using a camera and some milk.