Noise removal vs concentric zone model conversion question

[GMahmoud]

Hello @seungjae24

After diving in your patchwork implementation, I found some inconsistencies that make me wonder whether in some scenario swe can get duplicated point in the output point cloud.

I see at first the reflected_noise_removal is adding noisy points to cloud_non_ground_
Then in a second time the pc2czm is pushing points outside of zones to cloud_non_ground_.

How you can be sure that noisy points will not be duplicated with zone model conversion ? 🤔

Thanks again for you great work.

Cheers ! 😄

[seungjae24]

Hi @GMahmoud, thank you for interest on our work.

Actually, z components of noisy points, which are filtered out at the reflected_noise_removal function, are manually set to the value of std::numeric_limits::min(), which is not expected in usual point cloud data. Then, in the concentric zone model division, there is a conditioned part for points with the value std::numeric_limits::min() to skip the for loop, which guarantee noisy points are neglected in further procedures.

Honestly, the code looks somewhat weakly constrained and needs some tight prohibition for noisy points instead of using the specific value as std::numeric_limits::min(). But, it is true that the original code was also designed to skip noisy points in further procedures without any duplication.

Thank you :)

[GMahmoud]

Ok perfect !
My bad I missed this oneline
Thanks, no issue then 😄