What was accomplished

●       The software team evaluated several approaches for generating an aerial map from onboard imagery, comparing feature-based image stitching (using OpenCV's ORB feature matching and homography) against streaming then doing it in the ground station or recording it in the sd card found in the camera.

●       A lightweight, GPS-tagged image-stitching approach was selected as the baseline mapping method, offering faster processing suitable for near-real-time use during competition missions.

●       An initial proof-of-concept script was built to stitch a small set of sample aerial images into a single composite map, geo-referenced using logged GPS coordinates.

Challenges and solutions

The main challenge was balancing mapping accuracy against processing speed, since a full photogrammetry pipeline produces higher-quality maps but takes far too long to run for competition timelines. The team resolved this by adopting the lighter ORB-based stitching method as a first working baseline, with plans to optimize accuracy later once the pipeline is functional end-to-end.

Next steps plan

Begin training and tuning the YOLOv8 detection model on the competition target dataset.