Orthomosaic generation - Mesh vs DEM

[SimonBrown]

I'm now generating orthomosaics from every model I produce. The ability to review an underwater site in the comfort of your own home (and not get wet) is a great way of investigating what is on the seabed.

Example:

http://deep3d.co.uk/2017/08/23/bolt-bay-p47-d-crash-site/

When generating an orthomosaic there are two choices as to what to base the results on:

Mesh
DEM

I'm running a very large orthomosaic now (5 acres of shipwreck - 1 pix = 7.5mm) otherwise I would test it, but what are the differences? Any advantages over using the DEM vs the Mesh?

I am guessing it would be possible to generate a DEM from the sparse cloud and then derive the orthophoto from that, missing out the dense cloud and mesh, but would be interested to hear the experiences of others?

[JMR]

Ortho from DEM uually gives better quality and less artifacts near buildings.
DEM from sparse usally gives poor quality model. Keep in mind that points in the sparse cloud are not linked to important 3D features, but to pixel clusters that are good candidates for being found in several overlapping photos... nothing really linked to shape but with appearance. Analyse the blue key points superimposed in your photos and ask yourself if they can make a proper elevation model in your case.
I'd rather use a low qualtity dense cloud better than the sparse cloud in most cases unless the terrain is mostly flat.
Best regards.

[SimonBrown]

Thanks JMR - a very good explanation.

[SAV]

Hi SimonBrown,

There are only a few special cases where I would prefer the mesh over the DEM to compute my orthomosaic.

95% of the time I would use the DEM.

Why?

1. The DEM has generally a much higher resolution than your mesh (because the mesh is computational expensive to compute), hence the orthomosaic will show less 'artefacts' if it is based on the DEM and will also be more accurately corrected for distortion. This is especially important for geometrically complex scenes.

2. The DEM is much quicker to compute because it is less expensive from a computational point of view compared to a high quality mesh.

I hope that helps.

Regards,
SAV

I'm now generating orthomosaics from every model I produce. The ability to review an underwater site in the comfort of your own home (and not get wet) is a great way of investigating what is on the seabed.

Example:

http://deep3d.co.uk/2017/08/23/bolt-bay-p47-d-crash-site/

When generating an orthomosaic there are two choices as to what to base the results on:

Mesh
DEM

I'm running a very large orthomosaic now (5 acres of shipwreck - 1 pix = 7.5mm) otherwise I would test it, but what are the differences? Any advantages over using the DEM vs the Mesh?

I am guessing it would be possible to generate a DEM from the sparse cloud and then derive the orthophoto from that, missing out the dense cloud and mesh, but would be interested to hear the experiences of others?