Drone orthomosaic pipeline

What this project is

A drone flight leaves you with a pile of overlapping photos, and on their own they are hard to use. I built a pipeline that takes those photos, matches them to each other using epipolar geometry, and stitches them into one clean top-down map, called an orthomosaic.

The finished maps are saved to S3, so the client can open them whenever they need. It works well for farms, building sites and land surveys.

What needed solving

Drone photos overlap and are taken from slightly different positions and angles, so they cannot simply be laid side by side. Farmers, site managers and surveyors want one accurate top-down view of the whole area, not hundreds of separate pictures.

What I did

• Built the pipeline end to end, from raw drone photos to a finished orthomosaic.
• Used epipolar geometry to match overlapping images to each other.
• Set up S3-backed storage so the client can open finished maps at any time.

How it works

1. Collect

   Take the overlapping photos from a drone flight over the area.

2. Match

   Find the same ground points in neighbouring photos, and relate each pair of views with epipolar geometry.

3. Stitch

   Combine the matched photos into one clean top-down map, the orthomosaic.

4. Store

   Save the finished map to S3, where the client can open it anytime.

Epipolar geometry limits where a point from one photo can appear in another. That makes matching faster and cuts false matches between overlapping images.

Tools and methods

What came out of it

What the pipeline delivers:

• One clean top-down map built from many overlapping drone photos.
• Finished maps stored in S3, so the client can open them anytime.
• Works for farms, building sites and land surveys.