Penerapan You Only Look Once dan DeepSORT untuk Deteksi Plat Nomor Kendaraan

Hidayat, Firhat
Billy, Natanael
Permana, Nicholas Russel
Hariady, Matthew Evans

Jurnal Telematika
Vol. 19 No. 2 (2024)

Number plate detection is essential in traffic monitoring, law enforcement, and intelligent transport systems. However, existing methods still have difficulty accurately tracking vehicles in heavy traffic conditions. This study addresses this by combining the YOLOv8 detection model and DeepSORT tracking. Using 453 images from Kaggle, this study analyses the effect of batch size variation and an epoch on model performance. The best model achieved 95.5% precision, 95.1% recall, 98.7% mAP50, and 64.5% mAP95. The integration of YOLOv8 and DeepSORT can improve tracking consistency, reduce ID switching errors, and increase the reliability of the automatic number plate recognition system.