I’m trying to improve the depth accuracy of a 3D vision setup on a small mobile robot. I’m using a structured-light type depth camera, and I noticed some variation in Z-distance when scanning objects with shiny surfaces.Any suggestions on how to reduce reflection errors or smooth the point cloud data?

 

Infrared lighting. I had the same problem measuring solder height with 3D camera - if the solder was too shiny it was detected as a hole. I did not try infrared lighting to be honest, I had to solve the issue with the vision tools of the camera but when I was doing the research for the problem the conclusion was that the lighting for shiny objects should be infrared.

 

Have you tried asking the manufacturer of the camera? They should know the best way to deal with a host of issues.

 

MOST of the articles that I have read about solving distance resolving issues used two cameras spaced some distance apart

 

and this is no exception... shape of the camera is a sign. and look at the front shows there are several features including two that are likely camera lenses

 

Any suggestions on how to reduce reflection errors or smooth the point cloud data?

do talk to the manufacturer. every vision system needs tuning and correct lighting is part of it. there must be settings to control it. also pretty sure that manufacturer would have something to deal with it (exposure time etc). another is to use diffuse light. perhaps placing frosted piece of plastic or glass in front of light source is all that is needed. or adding additional light sources... or placing shade above the area so room light etc does not interfere with the measurement.