Z-accuracy and spatial noise

[njezersek]

Proof of concept for testing z-accuracy and spatial noise.

[saching13]

@njezersek can you remove binary files and json files. Add them to .gitignore in the sub folder

[alex-luxonis]

To not introduce the binary files into the repo history (that would increase its size), rewrite the branch history to remove the file additions, or, easier, use 'squash' when merging:

[saching13]

@njezersek Can you make a small video of usage and share it here ?

[saching13]

I haven't checked the math logic yet. Will do it after the following updates. In addition to above comments, it looks like you are testing two devices individually.

What we need to do is modify it to use the results from other device as ground truth.

1. Create the Base truth Wall to adjust the angle of the devices.
2. Create the Ground Turth Plane using the secondary device.
3. Use that ground Truth to estimate the Z-accuracy.
4. Fill rate, Plane FIT RMSE and Sub-Pixel RMSE stays same since they are Spatial Noise Estimators.

[njezersek]

Here is a screen recording demonstrating the usage (I had some problems with the screen recorder, so I recorded the screen with my phone)

Steps taken:

• Select ROI on both cameras
• Selected the Astra cam with the 2 key
• Set ground truth from the Astra cam measurement (g key)
• Selected the OAK cam with the 1 key
• Fit the plane to the OAK's point cloud (f key)
• Run the tests (t key)

https://user-images.githubusercontent.com/10494687/202206342-58c7b864-ae3b-4a47-a0d7-d81f4511385e.mp4

[saching13]

• Set ground truth from the Astra cam measurement (g key)

is this done using avg of the region selected ?

And using the astra pro as GT is optional right ?