Semester of Graduation

Fall 2023


Master of Science (MS)


School of Renewable Natural Resources

Document Type



Understanding how waterfowl respond to habitat restoration and management activities is crucial for evaluating and refining conservation delivery programs. However, site-specific waterfowl monitoring is challenging, especially in heavily forested systems such as the Mississippi Alluvial Valley (MAV)—a primary wintering region for ducks in North America. I hypothesized that using uncrewed aerial vehicles (UAVs) coupled with deep learning-based methods for object detection would provide an efficient and effective means for surveying non-breeding waterfowl on difficult-to-access restored wetland sites. Accordingly, during the winters of 2021 and 2022, I surveyed wetland restoration easements in the MAV using a UAV equipped with a dual thermal-visible high-resolution sensor to collect 2,360 optical images of non-breeding waterfowl. I then developed, optimized, and trained a SingleShot MultiBox Detection (SSD) object detection model with a deep convolutional neural network (VGG16) backbone to locate and identify eight different duck species in the UAV imagery. The final model achieved a total mean average precision and recall of 99.1% and 82.9%, respectively, after only 45 training epochs. The individual species class precision ranged from 65.3 to 86.1%, while the species class recall ranged from 69.7 to 88.6%. This study demonstrates the promise of UAV-based surveys for effectively surveying non-breeding waterfowl in structurally complex and difficult-to-access habitats and, additionally, provides a functional deep learning-based object detection framework for automated detection of non-breeding waterfowl in UAV imagery. This framework can be used to provide managers with an efficient and cost-effective means to count waterfowl on project sites—thereby improving their capacity to evaluate waterfowl response to wetland restoration efforts.



Committee Chair

Ringelman, Kevin M.