Photo alignment

Create a new project file for the zone. Name the file after the zone, for example: zone-4.psx.

Add a chunk, then rename it raw sparse cloud.

This folder will only hold the project’s raw data.

Reference system

Select the chunk > right-click > Reference settings
Coordinate System: Local Coordinates (ft)

Add the photos

Add all the photos for the zone.

Alignment

The photo alignment step is partly defined in a batch workflow file, 1. Align cameras.xml. It takes care of the following steps:

Runs a Python script that rates image quality and disables the images scoring below 0.1
Runs the photo alignment task with Accuracy set to Highest.

How to

Select the chunk(s) to align
Workflow > Batch Process...
Load...
config/workflows/1. Align cameras.xml

Unaligned photos

…

Additional tasks

Fix the object’s orientation on all 3 axes.

Adjust the region’s orientation to match the object’s. Resize the region so it covers the whole point cloud.

Aberrations

Go over the model looking for badly aligned parts. These parts can be realigned like this:

Select the points of the badly aligned part
Right click > Filter photos by Tie points
Show the cameras (Show cameras)
Select all the cameras
In the chunk, find the group(s) of selected images
To make sure you catch every image involved, you can select the neighbouring images and see which camera is highlighted in the model
Select all the images involved
Right-click > Reset Camera Alignment
Right-click > Align Selected Cameras

You can also disable the cameras if realigning doesn’t fix the problem.

Optimization

Make a backup copy of the chunk before the optimization steps:

First chunk: [Chunk name] - RAW SPARSE
Second chunk: [Chunk name]

Tie Points optimisation

After each step, run a data optimization (menu Tools > Optimize Camera Alignment), checking the f, k1, k2, k3, cx & cy, p1, p2 parameters and additional corrections.

Model > Gradual Selection

Reconstruction Uncertainty: <= 15 (recommended: 10, max 25% of points removed at a time)
Tools > Optimize cameras..., then Ok
Projection Accuracy: 5 (recommended: 2 or 3, max 50% of points removed at a time)
Reprojection Error: 0.4 (10% of total points)

Tie Point Accuracy: 0.1 (max 1)