Mesh it up – Depth Maps vs Dense Cloud

With the release of Metashape 1.5 we get a choice with the route to mesh creation:

  • Depth Maps
  • Dense Cloud

For some purposes we don’t really need a mesh, the dense cloud is just a series of defined points in 3D space but can have GIS information and scale applied. But for most uses – animation, CGI, 3D printing etc – its the mesh we want.

Much of the testing has been performed on the torpedo engine. Its a handy dataset, very repeatable and if I want to check/verify the 3D world with reality I only need go to my garage. Here’s a reminder of the model:

The torpedo motor currently undergoing restoration preserved here digitally.

Before delving into the detail, here’s a quick reminder about one of Metashape’s tools:

Metashape has a very handy tool for capturing views – ideal for quick grabs of everything from the model to the DEM or snippets of the ortho photo.

Dense Cloud Mesh

Lets be honest; sometimes generating the dense cloud can take days. The more images we have in the set, the longer the process. However, one of the causes of excessive processing time is excessive overlap. Given the time, trouble and risk of diving its little wonder those who practice underwater photogrammetry can be more than slightly paranoid about missing ‘the one’ image that prevents two chunks from working as one. But multiple overlapping images do increase processing time, and images can be disabled if needed. The results from one section of the dense cloud:

Mesh derived from dense cloud. For reference, the diameter of the fuel pipes (centre) are apex 6mm.

Depth Map Mesh

When generating a mesh with the depth maps option we skip the dense cloud – no 3D point cloud is generated and will exist. However, the time taken to get from sparse cloud to mesh improves and the resulting mesh is seen below:

The depth map derived mesh. The process filled gaps better than the point cloud and the definition of the fuel pipes (centre) looks clearer.

Which Method?

Both have their benefits and downsides. Getting to the mesh via a dense cloud takes longer, but the point cloud itself can have uses. Anecdotally, it looks like depth maps extract details from thin areas (like pipes) and in other testing the buoy line (10mm rope), thin light guards and electrical wire were clearly defined on the Saracen Armoured Car:

The light grille (lh), the buoy line (centre top) and electrical wire (rh, draped over the wheel) were more clearly defined when compared to dense cloud derived mesh.

The full model:

Final reminder: after alignment we can make a choice and choose depth maps or dense cloud route to the mesh. And if we don’t like it, the alignment is reusable either way. All the models here have been derived from single alignments. Depth Maps are just another tool in the kit and a very welcome one at that.

1 thought on “Mesh it up – Depth Maps vs Dense Cloud

Comments are closed.