Penguin colony georegistration using camera pose estimation and phototourism

Area of Science:

• Ecology and conservation biology
• Computer vision and photogrammetry
• Remote sensing and geospatial analysis

Background:

• Satellite and aerial imagery are crucial for wildlife monitoring but limited in remote regions.
• Ecotourism generates vast, unstructured photo data difficult to utilize for scientific purposes.
• Phototourism, a computer vision technique, reconstructs 3D scenes from diverse, unstructured photographs.

Purpose of the Study:

• To adapt phototourism for mapping Antarctic penguin populations using ecotourism photographs.
• To develop a semi-automated pipeline for georegistering ground-based photographs with environmental data.
• To demonstrate the utility of this approach for wildlife and habitat monitoring in challenging environments.

Main Methods:

• Utilized phototourism to reconstruct georeferenced 3D scenes from ground photographs.
• Developed a pipeline aligning and registering images using digital elevation models (DEM) and satellite imagery.
• Employed the Segment Anything Model (SAM) for interactive identification and segmentation of penguin colonies.

Main Results:

• Successfully created a textured 3D mesh from DEM and satellite data.
• Estimated camera poses to align ground photographs with the 3D mesh.
• Registered segmented penguin colony areas onto the 3D mesh, yielding detailed representations.

Conclusions:

• The phototourism pipeline offers a straightforward, effective tool for georegistering ad-hoc photographs in natural landscapes.
• The method shows promise for monitoring wildlife populations, such as Antarctic penguins, and habitats.
• Potential applications extend to monitoring environmental changes like glacial retreat.