Towards Automatic Burrow Detection for Sustainable River Levees

Burrows are tunnels or holes excavated into the ground by certain types of animals, to be used as habitation or temporary refuge, or as a by-product of their locomotion. Burrows provide a form of shelter against predation and exposure to the elements, and can be found in nearly every biome and among various biological interaction types. River bank burrowing weakens the soil structure, increases the risk of erosion, and may lead to bank retreat and landslides. Currently, burrow watching, mapping, and prevention are human-only activities, and there are no conventional data or information systems designed for this purpose. In this paper, we design, implement, and test a novel AI-based solution that, starting with drone-acquired imagery, allows the user to automatically identify and map potentially dangerous burrows in the target area, and lays the basis for the digitization and systematic conservation of such information, to be later used for intervention and planning. Our solution contributes to the environmental sustainability of rivers, especially close to densely populated areas.