Spikes, twisted edges and waves in pointcloud [F200]

[viktorku]

• Partially reflective materials cause spikes (peaks and valleys) in resulting pointcloud.
• Flat surfaces are often projected with wavy/dancing points (varies with distance from camera) and even more often the same flat surfaces have twisted (upwards) edges.

Example (average distance from camera ~30cm):

By twiddling and adjusting all parameters, best results for getting rid of the spikes gave drastic reduction of the laser power which also reduces the pointcloud size, and effectively renders it useless (even then the spikes and twisted edges are not completely eliminated).

Is this an expected behavior or something worthy to resolve?

Another example of twisted edges (37cm average distance):

Note: Screenshots are made after camera was tested and running (non-continuously) for several hours; I observed better average results than in the first ~hour after a cold start.

PS: While running cpp-config-ui I ran into TemperatureControlLoop error (#26 but only on one camera; also had it couple of times when running cpp-tutorial-3-pointcloud) and one-time irreproducible timed_mutex::try_lock_for(...) timed out exception which didn't freeze neither view.

More angles from the first example:

[ddiakopoulos]

Glad you documented this investigation of the F200. These artifacts are unavoidable on account of the core coded light approach of the F200 (and SR300). Laser power does have a direct impact, but it is one of the very few parameters a user can tune to mitigate issues found with reflective surfaces. Unfortunately there is little we can do from a software API point of view.

I'll follow up about the TemperatureControlLoop in a separate issue.