Deep learning framework for barcode localization and decoding using simulated UAV imagery

Automated warehouse stock tracking is becoming increasingly important for improving logistics and reducing manual errors. Unmanned Aerial Vehicles (UAVs) offer a promising solution by enabling automated barcode scanning from above. However, challenges such as poor lighting, shadows, and partial occlusions still limit the reliability of real-time barcode detection and decoding. This research presents a deep learning framework evaluated on simulated UAV imagery for barcode inventory management. The proposed system uses the YOLOv8 object detection model to accurately localize both 1D and 2D barcodes in images captured from a UAV perspective. With a mean Average Precision (mAP) of 92.4%, the model demonstrates strong performance even in complex warehouse conditions. Once localized, the barcode regions are decoded using OpenCV's barcode module. The extracted data, including product ID and quantity, is then automatically updated into a MySQL database to simulate real-time stock updates. Although tested using simulated aerial imagery, the system is designed to be drone-ready with minimal adjustments. This modular approach shows potential for real-world UAV deployment and contributes to reducing human effort and errors in inventory tracking.