Introduction to Photogrammetry

Taught by Trevor Coates

Course Number:
Software Version:
Original Run Date:
July 2014 
7 hours 51 minutes 
This course uses photogrammetry reconstruction techniques to create detailed object mesh and scene models from photographs. Aimed at matte painters, scene designers and previz artists. Photogrammetry can be thought of as an inexpensive alternative to LIDAR. In this course, Photoscan will be the reconstruction tool of choice combining with Maya, Matte Painters Toolkit, Mari, Mudbox and zBrush with support from PhotoShop, Lightbrush, PfTrack and Syntheyes.

There is no one single approach that will solve for all shots. The course will range across a number of tools exposing possible workflow approaches and useful tool combinations that naturally flow from Photoscan. In this way you will be provided with the knowledge to formulate your own approach.

We’ll be looking at outdoor object, scene and aerial scans as we step our way through the entire workflow from photographs to render. All the while working to get the best out of Photoscan. Including a look into extracting Albedo from the source images producing textures for use in physically based rendering techniques via Lightbrush.

Using a number of example data sets the course exposes many of Photoscan’s export options. The most common being mesh, textures, cameras and point clouds and explore their uses outside of Photoscan. The principles in this course are not restricted to Maya.

Trevor is a Generalist Digital Artist who has been in the TVC, 3D, visual effects and motion graphics industry since 2000. With prior experience in graphic design followed by multimedia with five years teaching at The CGC now Quantum College. Currently freelance in Sydney working largely on TVCs and realtime projects.
Introduction to Photogrammetry
Watch our overview of the course

Class Listing

Class 1

Introduces the course discussing what is to be developed throughout the classes. A run through of the six main course projects and what they entail. Including a demo style run through of project one Busbys Bore the content from class two. Ending with a look into camera specifications and a few issues around shooting for Photoscan. Purpose - show course content and develop an introductory appreciation of the process.

Class 2

Taking the files demonstrated in class one Busbys Bore many steps deeper. Understanding photographs as data set, shooting patterns, resolution, image formats and qualifying images for use in Photoscan and the effects on processing speed. Working through image importation and processing, the chain of steps to model, texture creation and on to export to Zbrush and automated retopology. The result reimported into Photoscan for texture re-projection and export to FBX for use in Maya.
Purpose - develop a deeper understanding of Photoscan, decimation and workflow with the Zbrush Remesher tool.

Class 3

Using the Federation Dome project we again look at shooting patterns and work through at a deeper level from images to textured model and export via FBX to Maya. This time introducing camera projection using Matte Painters Toolkit. Along the way image alignment control and refinement pivotal to modelling success in Photoscan is covered in depth.
Purpose - develop a deeper understanding of image alignment and workflow introducing Matte Painters Toolkit.

Class 4

This time the Centennial Fountain project is used to continue looking into shooting patterns, site issues and work through at a deeper level from images to textured model and export via FBX to Maya. Introducing Lightbrush a method for reflectance texture creation for physically based rendering.
Introduces Photoscan's batch processing and round tripping back to Photoscan from Maya for texture re-projection.
Purpose - understanding of Photoscan's batch processing and using Matte Painter's Toolkit as a modelling assistant and introducing Lightbrush.

Class 5

Mother's Day Footy statue. Each project has its own set of requirements and so again shooting patterns are discussed. Working through the Photoscan model creation this time looking into ways of fixing the meshing problems. Preparation for Mari its UDIMS and projectors. Working through direct export from Photoscan to Mari and from Maya to Mari.
Purpose - variations on Photoscan UV creation introducing Mari into the workflow mix.

Class 6

Chopper Aerial Photography - we will be working with some aerial photography shot with a DSLR rigged to a quad copter. Once the frames have been extracted from the footage they are taken through the usual steps in Photoscan. The dense point cloud and model are exported from Photoscan and used in The Pixel Farm's PfTrack software. The footage is tracked then aligned with the point cloud the sum of that alignment is then used to align both the imported mesh and the extra mesh created in PfTrack. The aligned camera and mesh are then exported out to Maya and the scene recreated.
Purpose - preparing video footage for model creation in Photoscan. Using point cloud data to align the tracked original footage to the Photoscan model using PfTrack for reassembly in Maya.

Class 7

We tackle a fairly common problem with camera alignment using SynthEyes. From time to time a camera position or focal length change can throw the alignment of one camera out which can in turn affect the alignment of the following cameras. Using SynthEyes to manually pre-align the cameras fixes this problem.

Class 8

We start out with reference to a forum question regarding how to colour balance images in preparation for use in Photoscan. Followed by an introductory look at PhotoScan's chunks in preparation for the Centennial Drive project.

Class 9

We begin with the final visit to Zbrush to use Dynamesh to fill meshing holes followed by a reference to a forum question regarding shooting and processing interior spaces in Photoscan. Then onto the preparation set ups for the Centennial Drive project.

Class 10

The Centennial Drive project is finished off with a further look into how chunking can assist with managing large data sets. Then the course is finished off with a look into shooting for a working with a small object data set based around a sand shoe.