This is an interesting paper on laser triangulation sensors. However, it appears to be old and may be a little bit outdated.
20160703/https://www.ogpnet.com/Assets/uploads/files/OGP/TriangulationSensors.pdf
One thing particular about this research paper is that it comments a lot on 3D scanners that use a position sensitive device (PSD) for the image sensor.
So, are these position-sensitive devices 1D only and must they be lined up in an array to sense an image? Well, let’s see what Wikipedia has to say. Wikipedia has some information on this, but it’s not very interesting.
- Particularly, it became apparent to me that you could use a PSD sensitive to infrared light as the trackpad on a laptop computer.
20170319/https://en.wikipedia.org/wiki/Position_sensitive_device
So, it turns out that there are actually 2D position-sensitive devices. So, no need to build an array if you are only using a point-laser light source, which is what the paper appears to describe is recommended for such 3D scanners.
Also, the paper notes that a small spot size is needed for good resolution on such devices since they are threshold-driven. Note that this is rather different than the requirement.
So, if you do want to do swept-line laser triangulation using such a device, it has to be done using an array of 1D position sensitive devices. Otherwise, the PSD will get confused. And, even then, this can only be done under certain conditions when the laser line appears in the correct direction so that there is only one dot apparent on each 1D PSD. Even then, you still won’t get as good image quality as the 3D scanner is threshold-driven. So yes, you know what the Stanford Computer Graphics Lab research paper was saying about that? It results in poor quality 3D scans with various aliasing artifacts due to material property variations.
Wheras, in the case of CCD/CMOS image sensor based devices, you can use phase-shift profiliometry. This allows you to not only get better results, but it allows you to get results limited by the resolution of the imaging device, not that of the laser line thickness. Also, even without perfectly thin laser lines, you can still get better results due to the use of multisampling for each single point in the picture.
To summarize, although at first, it may appear that the CCD/CMOS method is inferior, it turns out that there are lots of advantages of using the particular method, not just in level of detail but also in ease of use by being able to use a laser line rather than a laser dot. In the end, the cheaper solution using commodity image sensors actually gives better results.