Problem to compute normals

[Orion]

Hi, I been using the program to work with cloud points of a tunnel excavation and creating a mesh with Poisson recon, the problem is no matter what configuration I use to compute the normals they always turn to be blacks dots on the cloud. Also the cloud is not in an singular axis, but planar is the best option for the tipe of surface I'm working on. I can create the mesh anyway, so I don't know how this might be affecting the mesh creation process.
I put images of how the cloud point looks.

The cloud of the front of the tunnel excavation

Captura de pantalla 2026-03-04 113733.png (1.3 MiB) Viewed 4257 times

Computed normals in the front

Captura de pantalla 2026-03-04 113401.png (188.2 KiB) Viewed 4257 times

Computed normals in the back

Captura de pantalla 2026-03-04 114011.png (43.14 KiB) Viewed 4257 times

Mesh whit RGB colors

Captura de pantalla 2026-03-04 115345.png (2 MiB) Viewed 4257 times

[daniel]

For the 2nd and 3rd images: aren't the normal simply inverted? That would explain why you have a clean model in the end.

And have you tried the Minimum Spanning Tree algorithm to (try to) set the orientations in a more dynamic way?