Sistem Pengenalan Pelat Nomor Menggunakan Metode YOLOv8 Dan PaddleOCR Pada Sistem Parkir Pintar

Narawangsa, Mochammad Daffa (2025) Sistem Pengenalan Pelat Nomor Menggunakan Metode YOLOv8 Dan PaddleOCR Pada Sistem Parkir Pintar. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040211054-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Sistem parkir berbasis RFID di Fakultas Vokasi ITS memunculkan antrean karena proses validasi yang memerlukan kontak fisik dan seringnya gangguan pembacaan sensor. Penelitian ini mengembangkan sistem Automatic License Pelate Recognition (ALPR) yang lebih cepat dan praktis dengan memadukan YOLOv8 untuk pendeteksian pelat dan PaddleOCR untuk pembacaan karakter. Dataset berisi 4,035 citra pelat nomor didapatkan dari rekaman lapangan serta Roboflow Universe, lalu diperbanyak melalui augmentasi rotasi ±15° dan variasi kecerahan guna meningkatkan keragaman dan mencegah overfitting. Model YOLOv8 dilatih hingga 100 epoch pada resolusi 640×640 piksel, sedangkan PaddleOCR menggunakan SVTR untuk mengidentifikasi karakter. Evaluasi pada 153 karakter uji menunjukkan precision 98,69 %, recall 98,69 %, dan F1-score 98,69 %. Sistem juga diuji pada jarak 50–200 cm dan intensitas cahaya 2–948 lux, akurasi optimal tercapai di atas 182 lux, menurun drastis pada 2 lux sehingga dibutuhkan pencahayaan tambahan. Pada sistem barrier gate, mikrokontroler ESP32 menggerakkan servo RDS3235 untuk membuka (90°) dan menutup (0°) gate dengan akurat. Hasil ini membuktikan bahwa integrasi deteksi, pengenalan, validasi database, dan barrier gate dapat bekerja secara end-to-end dalam skenario lapangan. kombinasi YOLOv8 dan PaddleOCR memenuhi kebutuhan untuk akurasi yang tinggi, sehingga layak diterapkan sebagai solusi parkir pintar di kampus maupun lingkungan serupa.
================================================================================================================================
RFID‑based parking at the ITS Vocational Faculty suffers from long queues because it still requires physical contact for card validation and is prone to reader failures. To overcome these bottlenecks, this study builds a fully contact-less Automatic License Plate Recognition (ALPR) system that couples YOLOv8 for plate detection with PaddleOCR for character recognition. A dataset of 4,035 motorcycle‑plate images captured on campus and sourced from Roboflow Universe was augmented through ±15° rotations and brightness shifts to enrich variability and curb overfitting. YOLOv8 was trained for 100 epochs at 640 × 640 px, while PaddleOCR handled character inference with SVTR. On a 153‑character test set the system achieved 98.69 % precision, 98.69 % recall, and a 98.69 % F1‑score. Field trials across 50 – 200 cm stand‑off distances and illuminance levels of 2 – 948 lux revealed that accuracy remains optimal above 182 lux but degrades sharply at 2 lux, indicating the need for supplementary lighting in dim environments. An ESP32 microcontroller reliably actuated an RDS3235 servo to open (90°) and close (0°) the barrier gate, completing the end‑to‑end workflow of detection, recognition, database validation, and gate control. These results demonstrate that integrating YOLOv8 and PaddleOCR delivers the high accuracy required for practical smart‑parking deployments. The proposed ALPR‑based gate system eliminates physical contact, shortens validation time, and is well‑suited for campuses and similar facilities seeking to modernize their access control infrastructure.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	ALPR, CNN, OCR, Smart Parking System, YOLOv8.
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76 Computer software > QA76.8 Microprocessor T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T59.7 Human-machine systems. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.585 TCP/IP (Computer network protocol)
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Mochammad Daffa Narawangsa
Date Deposited:	07 Aug 2025 06:39
Last Modified:	07 Aug 2025 06:39
URI:	http://repository.its.ac.id/id/eprint/127940

Actions (login required)

View Item