Chemical plants present hazardous environments for manual safety inspections, driving the need for automated solutions. This paper presents an autonomous drone navigation framework using Quick Response (QR) code markers and deep learning to enhance safety inspection processes in chemical pilot plants. The proposed system integrates a lightweight YOLOv3 object detection model for real-time QR code recognition with an unmanned aerial vehicle (UAV) to navigate complex industrial layouts without human intervention. A DJI Tello micro-drone was deployed in a multi-level pilot plant (Hydrotreater Unit) using strategically placed QR code waypoints encoding navigation instructions. Experimental results demonstrate that the model achieved perfect QR code detection (average precision 1.00) on static imagery and $88 \%$ precision on more challenging dynamic footage, indicating robust recognition capability. The autonomous drone successfully followed the QR-guided route through the pilot plant, significantly reducing human exposure to hazardous areas.
Deep Learning Enabled Autonomous Drone Navigation for Safety Inspections in Chemical Plants
27.06.2025
405296 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch