Leuze LPS36/HI model. Minimum resolution is 0.1 mm
... 0.1 resolution in space range 4x3х2м. ...
0.1m or 0.1 mm?
For the mm: think you will not find a sensor with resolution of 0.1mm for a range of 2m (at least not for low cost).
So will have to make small partial scans and 'glue' them together.
How large are the models? Because robot accuracy is not going to get you 0.1mm over 1000mm ranges -- heck, not even 100mm distances. So laser-line sweeping is probably not doable, not without ~$150000 in hardware and software.
You'll need something like a photogrammetry solution for hand-held 3D scanning, that does not require a laser tracker to monitor the scanner position. The key is to use a hardware&software suite that can "stitch" different photos together while tolerating potentially substantial errors in camera position.
Capturing 3D point clouds over mulit-meter volumes is going to require pretty major software and memory, though.
A robot mounted Ctrack laser scanner will do what you need, but not cheap unfortunately
How large are the models? Because robot accuracy is not going to get you 0.1mm over 1000mm ranges
The main task that the scanner has to be used for is collecting main contours of car chassis base and other car outer body parts, such as headlights and so on.
The main task that the scanner has to be used for is collecting main contours of car chassis base and other car outer body parts, such as headlights and so on.
Well, my best suggestion is to try some of the open-source options and get a feel for what you need. I strongly suspect that, if you scan an entire car, you're going to end up needing (at minimum) a very powerful computer (top-end gaming or bitcoin miner rig) with a lot of storage and RAM. And be prepared for some very slow image processing.
I've used a Keyence XR-HT40 with an XG-X controller. These cost me in the low $30,000's
Very small field of view but great resolution. I believe image stitching is supported.
i used Keyence LJ profiler as a stationary sensor to scan products presented by robot. Different heads exist. One that I worked with has laser line about 150mm wide. As i recall resolution was some 2000 samples per line so choosing head with narrower field of view should give more detailed output. Not sure if the depth resolution will work for you. For example scanning coin surface was usually recognizable but fine details were not visible. This maybe because i only did one pass, so going slower, doing multiple passes or using averaging may produce better results. I wrote own code to interface to it and get the data. This sensor can generate quite a flood of data.
