Alitra

Library for ALIgnment and TRAnsformation between fixed coordinate frames. The transform is described by a translation and a homogeneous rotation.

Developed for transforming between the fixed local coordinate-frame and the asset-fixed coordinate-frame.

Installation

Installation from pip

pip install alitra

import alitra
help(alitra)

Image alignment (see Image alignment below) requires the optional image-alignment extra:

pip install alitra[image-alignment]

Installation from source

git clone https://github.com/equinor/alitra
cd alitra
uv sync --extra dev

You can test whether installation was successful with pytest

uv run pytest

Dependencies

The dependencies used for this package are listed in pyproject.toml and pinned in uv.lock. This ensures our builds are predictable and deterministic. This project uses uv for dependency management:

uv lock

To update the dependencies to the latest versions, run:

uv lock --upgrade

Image alignment

When two photos of the same scene are taken at different times, small differences in camera position mean a region-of-interest (ROI) polygon drawn on one photo no longer lines up with the same spot in the other. Alitra provides two functions to re-align a reference photo (and its ROI polygon) to a new source photo, so the polygon can be reused without redrawing it:

• align_two_images_translation_cv2: estimates a simple (dx, dy) pixel shift using phase correlation. Fast and robust, but only correct when the camera hasn't rotated or changed perspective between the two photos (e.g. a fixed camera with minor positional drift).
• align_two_images_orb_bf_cv2: detects ORB keypoints and estimates a full homography (rotation, scale and perspective change included). Handles a moved/rotated camera, but needs enough distinct visual features to match between the two photos.

Both functions return the ROI polygon transformed into the new photo's coordinates. If the transformed polygon has no overlap with the new photo at all, None is returned instead. If it partially overlaps, the returned coordinates are clamped to valid pixel indices.

Contributing

We welcome all kinds of contributions, including code, bug reports, issues, feature requests, and documentation. The preferred way of submitting a contribution is to either make an issue on GitHub or by forking the project on GitHub and making a pull request.

How to use

The tests in this repository can be used as examples of how to use the different models and functions. The test_example.py is a good place to start.