Automatic Waterfowl and Habitat Detection using Drone Imagery and Deep Learning

The integration of drone technology and Artificial Intelligence (AI) has opened up new possibilities for wildlife conservation and habitat monitoring. In this paper, we present a new system for efficiently and accurately analyzing waterfowl populations and classifying their habitats over large natural areas using drone imagery and deep learning (DL). Given a sequence of drone images captured by a drone along a flight path, the system utilizes customized deep learning models for waterfowl detection and counting, Meta’s SAM for image segmentation and customized deep learning models for segment classification, and ChatGPT to generate text-based survey reports. Several image overlap detection methods were developed and compared with. Our experimental results show accurate waterfowl and habitat detection results and improvement over previous work, providing efficient and accurate data analysis for wildlife conservation efforts.