Pix4d creates a mesh with poisson surface reconstruction.
The oriented point normal is very important when do poisson surface reconstruction.
The usual way to calculate points using nearby points is to calculate the wrong points for the building side. Therefore, It needs a normal orient process to modify the direction of this point
Generally, They use the viewpoint of a point cloud to adjust the direction of the normal, but Point clouds extracted from multi-images have a large number of view points.
How does pix4d solve this?
When the number of images is small, the model is fine.
But the number of images increases and the boundaries become wider, there is a problem with the shape of the model.
I want an answer please.