Tip 58: Denser Meshes in PhotoModeler’s MVS with a 2 Photo Min

Posted on Monday, March 31st, 2014

tip 58 blog img‘Multi-View Stereo’ is the new Dense Surface Model extraction algorithm introduced to PhotoModeler Scanner a few months ago. The ‘multi’ part of the name refers to the multiple photographs used to extract and confirm matching data. MVS and DSM are methods in photogrammetry for extracting a 3D point cloud (like a laser scanner) of a surface, object or scene.

To create a 3D point in the cloud, PhotoModeler’s MVS algorithm looks for corresponding matches across at least 3 photographs. This is a good minimum to ensure quality data. There are scenarios (mostly due to insufficient photo overlap) where data will be missing from the point cloud with a 3-photo minimum. That is, a point in the scene or on the object appears in only 2 photos and so MVS will not place it in the output cloud.

This tip video shows how to change the minimum number of MVS photos to 2. Doing so adds 3D point cloud data to areas missing in the example given (a stone wall). ¬†While 2-photo data can be noisier and of lower quality there are situations where photographs cannot be retaken and it is better to have some lower quality data, than to have none. The video also demonstrates an interesting review technique using PhotoModeler’s 3D Viewer frustum display to see areas lacking sufficient 3-photo overlap.

Watch the tip video below on Youtube.



©Copyright 2017 Eos Systems Inc.