A team of researchers from Stony Brook University has developed a method to analyze tourist photos for tracking ecological changes in Antarctica. The study, led by Heather Lynch, a professor at the Institute for Advanced Computational Science, focuses on penguin populations as indicators of environmental shifts due to climate change.
The technique uses computer vision to determine the location of penguins in tourist photographs. This is achieved by combining satellite imagery with 3D computerization methods to pinpoint where the camera was positioned when the photo was taken. The process also involves delineating the boundaries of penguin colonies using an AI model called the Segment Anything Model.
Lynch notes, “There are far more tourists in Antarctica than scientists, and virtually everybody has a camera in their pocket and many take photos of penguins.” The challenge lies in extracting information about penguin locations from these images without additional data.
The research addresses two main problems: segmentation and georeferencing. Segmentation involves identifying colony edges in photographs, while georeferencing determines geographic coordinates of penguins within those images. These tasks are complicated by Antarctica's lack of static features typically used for creating 3D models from 2D images.
By overlaying satellite images onto digital elevation models, researchers created a 3D representation of islands inhabited by penguins. They then used this model to determine camera positions and extract precise colony boundaries.
“In theory, this information gathered by the computational technique can be compared to other similarly processed images of the Antarctic to see how penguin colonies are changing over time,” says Lynch.
While traditional methods like satellite and aerial photography are common for monitoring Antarctic changes, they have limitations. Tourist photos could significantly increase available data for long-term environmental studies.
Despite promising results, challenges remain due to varying image quality and dynamic landscapes. The researchers describe their method as “a straightforward and effective tool for the georegistration of ad-hoc photos in natural landscapes.”
This interdisciplinary project involved collaboration among ecologists, computer scientists, mathematicians, and geologists at Stony Brook University.