Implementasi YOLOv11 dan Vision Language Model untuk Deteksi dan Pembacaan Plat Nomor Kendaraan

Susetio, Nabila Mutiara (2026) Implementasi YOLOv11 dan Vision Language Model untuk Deteksi dan Pembacaan Plat Nomor Kendaraan. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024221037-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Sistem pengenalan plat nomor kendaraan (Automatic License Plate Recognition/ALPR) memiliki peran penting dalam berbagai aplikasi, seperti sistem parkir otomatis, pengawasan lalu lintas, dan penegakan hukum. Penelitian ini berfokus pada pengembangan dan evaluasi sistem ALPR yang menggabungkan model deteksi objek dan Vision Language Model (VLM) untuk meningkatkan akurasi pembacaan plat nomor kendaraan. Pada tahap pertama, penelitian ini mengembangkan model deteksi plat nomor menggunakan YOLOv11 yang dilatih khusus untuk mendeteksi area plat nomor pada citra kendaraan. Model tersebut digunakan untuk menghasilkan bounding box yang kemudian dipotong (cropped) sebagai masukan bagi tahap berikutnya. Selanjutnya, penelitian ini mengevaluasi performa berbagai jenis VLM dalam membaca isi plat nomor dari ratusan sampel data. Evaluasi dilakukan menggunakan empat metrik utama, yaitu Character Error Rate (CER), Word Error Rate (WER), plate accuracy (rasio plat yang dibaca dengan benar terhadap seluruh data), serta latency atau kecepatan pemrosesan. Hasil evaluasi digunakan untuk mengidentifikasi VLM dengan performa terbaik dalam tugas pengenalan plat nomor. Setelah menemukan VLM yang memiliki performa paling unggul berdasarkan pengujian kuantitatif, penelitian ini menyusun sebuah pipeline ALPR yang mengintegrasikan YOLOv11 sebagai pendeteksi plat dan VLM terpilih sebagai pembaca karakter. Pipeline ini memungkinkan proses end-to-end: mulai dari deteksi plat nomor, pemotongan area plat, hingga pembacaan teks secara otomatis menggunakan VLM. Penelitian ini diharapkan dapat memberikan kontribusi pada pengembangan sistem ALPR yang lebih akurat, cepat, dan adaptif, serta menunjukkan potensi penggunaan VLM dalam tugas-tugas pengenalan teks berbasis visual di domain Computer Vision.
=========================================================================================================================================
The vehicle license plate recognition system (Automatic License Plate Recognition/ALPR) plays an important role in various applications, such as automated parking systems, traffic monitoring, and law enforcement. This study focuses on the development and evaluation of an ALPR system that combines object detection models and Vision Language Models (VLMs) to improve the accuracy of vehicle license plate reading. In the first stage, this study develops a license plate detection model using YOLOv11, specifically trained to detect the license plate area in vehicle images. The model is used to generate bounding boxes, which are then cropped as input for the next stage. Subsequently, this study evaluates the performance of various types of VLMs in reading the content of license plates from hundreds of data samples. The evaluation is conducted using four main metrics: Character Error Rate (CER), Word Error Rate (WER), plate accuracy (the ratio of correctly read plates to the total data), and latency or processing speed. The evaluation results are used to identify the VLM with the best performance in the license plate recognition task. After identifying the VLM with the best performance based on quantitative testing, this study constructs an ALPR pipeline that integrates YOLOv11 as the plate detector and the selected VLM as the character reader. This pipeline enables an end-to-end process: from license plate detection, cropping the plate area, to automatic text reading using the VLM. This study is expected to contribute to the development of more accurate, faster, and adaptive ALPR systems, as well as demonstrate the potential use of VLMs in visual text recognition tasks within the Computer Vision domain.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	ALPR, YOLOv11, Vision Language Model, deteksi objek, pembacaan platnomor, Character Error Rate, Word Error Rate, plate accuracy, latency ALPR, YOLOv11, Vision Language Model, object detection, license plate read- ing, Character Error Rate, Word Error Rate, plate accuracy, latency
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence
Divisions:	Faculty of Electrical Technology > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Nabila Mutiara Susetio
Date Deposited:	29 Jan 2026 05:36
Last Modified:	29 Jan 2026 07:43
URI:	http://repository.its.ac.id/id/eprint/130940

Actions (login required)

View Item