Currently I have a point cloud coming from a single depth camera (Microsoft Kinect for Azure). In the resulting point clouds there are of course occlussion and gaps. I wanted to create a mesh from this incomplete point cloud, but without creating triangles / vertices where nothing can be seen.
I tried Poisson Meshing, which resulted in the following files:
PCD is my original pointcloud
PLY is the resulting mesh
This is kind of bad, but I found no setting for the PCL library to NOT fill the gaps in between the original data. How can I approach this, or what is a better method?
Note: for the single camera, I could just do it with neighbouring pixels, but in the future there will be multiple cameras viewing the scene from different point of views, so I need some meshing algorithm that meshes all the data that is there, but ignores gaps in the point cloud.