2024 · Hardware / Robotics
Autonomous Drone Aerial Litter Mapping & Recovery System
A low-cost drone mapping system that spots litter from the air and plots it on a map for cleanup crews.
TVSEF Gold Medal · OES Excellence Award · Research Paper
The problem
Municipal cleanup effort gets allocated by intuition. A crew decides a park needs attention and walks whatever trail is easiest. Nobody has an accurate map of where illegal dump sites actually are because surveying large parks on foot costs more person-hours than the cleanup itself.
The constraint
Aerial computer vision is difficult: targets are tiny (bottles and cans occupy a few pixels from 15 meters up), ground textures (grass, gravel, shadows) are noisy, and detections are useless without accurate real-world GPS coordinates.
What I built
An F450 quadcopter pairing a Pixhawk flight controller (running ArduPilot for 400Hz real-time stabilization) with a Raspberry Pi 4B companion computer over hardware UART at 57600 baud. The Pi runs YOLOv8 aerial waste detection and correlates camera timestamps with flight telemetry to export georeferenced GeoJSON map layers for QGIS.
The outcome
Mapped a 50 m² survey in 5 minutes with 88% detection accuracy and automated GeoJSON exports. Won Gold at TVSEF, the OES Excellence Award, and published a formal research paper.