As part of a virtual production workflow, it is often required to create a digital version of a physical set, or to create a virtual version of a real-world location. These can be used in a multitude of ways:
- To scout camera positions, either in virtual reality, on a workstation or as part of a virtual camera session.
- To plan action sequences without needing access to the location
- In previsualization, to create animated storyboards or fully animated sequences.
- As an on-set reference tool, using Simulcam / real-time compositing.
- As final-picture assets filmed with ICVFX.
- As reference for post-production, showing set pieces, light placement etc.
Photogrammetry:
Photogrammetry is a process that uses overlapping still photographs or a sequential video file to extract a three-dimensional object. The input data can be captured with consumer grade cameras, and the image data is also used to texture the object, creating a high-resolution asset to be ingested into a digital workflow. Aerial Photogrammetry uses a drone-mounted camera to capture the images, resulting in a large-scale landscape mesh.
- Assets created in this way can not be used out-of-the-box for ICVFX. When photo-realistic assets are required, an Asset Artist will need to clean up and prep the object before it is used.
- Photogrammetry struggles with reflective objects, or featureless environments.
- A large number of images are required to calculate the most accurate result.
- The output object will be created at an arbitrary scale. An object of known size must be included in the scan to allow accurate scaling.
Image 1: An Aerial / Drone scan, showing camera positions
Image2: An object scan, showing camera positions and images
Hardware required:
A DSLR Stills camera / Drone Camera (depending on quality required and accessibility of location))
Software required:
Reality Capture / Agisoft Metashape / Autodesk ReCap
Photogrammetry can also be used to scan actors, although a different hardware setup is required (multiple synchronized static cameras, rather than one camera in multiple locations).
LIDAR Scanning:
LIDAR is an acronym for ‘light detection and ranging’ and uses custom hardware scanners to bounce laser pulses that create a three-dimensional map of an environment. Depending on the scanner used, this will either be a greyscale mesh object, or a textured / colored mesh.
Hardware required:
A LIDAR scanner. (FARO, Leica)
Software required:
FARO Scene / Scanner manufacturer software
Notes:
- Assets created in this way can not be used out-of-the-box for ICVFX. When photo-realistic assets are required, an Asset Artist will need to clean up and prep the object before it is used.
- LIDAR struggles with reflective surfaces, as it uses bounced light.
- Scans are ideally done with a clear set.
- Multiple static scans are created and merged in a post-processing step into a cohesive point cloud, which can then be meshed.
Image 1: A FARO LIDAR scanner
Image 2: Resulting LIDAR data
Hybrid Scanning:
If the highest accuracy is required, the best approach is a combination of the above. LIDAR scan data gives the highest scale and world-space accuracy, while adding high resolution images from a DSLR camera gives the highest texture accuracy. The data from both approaches can be combined using photogrammetry software and output as an accurate, high resolution textured asset.
If the asset will be used on camera, or absolute accuracy is required for technical visualization or detailed action planning, this is the recommended approach.