Automatic detection of birds in images acquired with remotely piloted aircraft for managing wildlife strikes to civil aircraft

Abstract

Wildlife strikes represent a significant threat to aviation safety and economics, as evidenced by notable incidents such as the emergency landing of US Airways Flight 1549 in 2009. The Federal Aviation Administration (FAA) acknowledges this growing concern, as evidenced by its report on Wildlife Strikes to Civil Aircraft in the United States from 1990 to 2019.To address the need for improved risk management, airports conduct Wildlife Hazard Assessments (WHA), a laborious process typically relying on visual identification by Qualified Airport Wildlife Biologists (QAWB). However, technological advancements, such as Remotely Piloted Aircraft (RPA) and machine learning (ML), offer promising solutions to enhance WHA efficiency. The latest model, YOLOv8, achieved a training accuracy of 81% in bird detection precision, and field testing yielded a 69.77% success rate.

Authors

• Oviedo, Maurycio Rodrigues Embry-Riddle Aeronautical University
• Mendonca, Flavio A. C., Ph.D. Embry-Riddle Aeronautical University
• Cabrera, Jose Embry-Riddle Aeronautical University
• McNall, Cole A Embry-Riddle Aeronautical University
• Ayres, Raymond Embry-Riddle Aeronautical University
• Barbosa, Ricardo Luís Embry-Riddle Aeronautical University
• Gallis, Rodrigo Embry-Riddle Aeronautical University

Keywords

• Automatic detection of animals in images
• UAV
• wildlife strikes to civil aircraft.
• Biology
• Ecology and Evolutionary Biology
• Forest Management
• Space Habitation and Life Support