A Bioinspired Framework for Person Detection and Tracking using Events and Frames on Flapping-Wing Aerial Robots
Raul Tapia, A. Ijjeh, J. Ramiro Martinez de Dios, Aníbal Ollero
Abstract
Flapping-wing aerial robots offer significant advantages over conventional multirotors, including lower noise signatures, higher energy efficiency, and enhanced maneuverability. Despite these benefits, their application in surveillance, particularly person detection and tracking, remains largely underexplored. This paper proposes a bioinspired framework for person detection and tracking, specifically designed for flapping-wing aerial robots. Drawing inspiration from the dual pathways in biological vision, our method integrates an event-by-event blob tracker with a more accurate but slower frame-based detector. The event-based tracker leverages the high temporal resolution and robustness to motion blur of event cameras, effectively compensating for the strong vibrations caused by the flapping strokes of these robots. The frame-based detection (implemented using a deep neural network) periodically corrects and enhances the event-based tracking estimates, globally achieving a balanced trade-off between accuracy, responsiveness, and computational cost. Evaluation with both multirotor and flapping-wing aerial robots validates the effectiveness and efficiency of the approach.
BibTeX
@inproceedings{iros2025_abioinspiredfram,
title = {A Bioinspired Framework for Person Detection and Tracking using Events and Frames on Flapping-Wing Aerial Robots},
author = {Raul Tapia and A. Ijjeh and J. Ramiro Martinez de Dios and Aníbal Ollero},
booktitle = {IROS 2025},
year = {2025}
}