Dyanti, Chata Trista (2024) Sistem Identifikasi Plat Nomor Kendaraan Menggunakan Metode Convolutional Neural Network Dalam Rangkaian Sistem Pengisian Tangki Distribusi. Diploma thesis, Institut Teknologi Sepuluh Nopember.
Text
2040201024_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (11MB) | Request a copy |
Abstract
PT. XYZ berkomitmen untuk menerapkan konsep Zero Waste dalam setiap produknya dengan mengolah limbah produksi menjadi pupuk cair yang dibeli oleh mitra vendor melalui sistem deposito. Namun, proses pencatatan identitas tangki vendor yang dilakukan secara manual menggunakan kertas yang menyebabkan adanya penumpukan limbah kertas. Untuk mengatasi masalah ini, diusulkan implementasi sistem identifikasi plat nomor kendaraan menggunakan metode Convolutional Neural Network (CNN) dengan algoritma YOLOv8. YOLOv8 memungkinkan deteksi plat nomor secara real-time dengan akurasi tinggi, sehingga mempercepat proses pencatatan dan mengurangi limbah kertas. Hasil pengujian menunjukkan bahwa model YOLOv8 dengan 100 epoch memberikan performa terbaik dengan menghasilkan presisi sebesar 91,90%, recall sebesar 64,14%, dan akurasi sebesar 64,14% dalam mendeteksi plat nomor dan menghasilkan presisi sebesar 88,68%, recall sebesar 88,73%, dan akurasi sebesar 99,93% dalam identifikasi karakter plat nomor. Namun, jumlah epoch dan jarak objek terhadap kamera mempengaruhi kinerja sistem, di mana penggunaan epoch yang berlebihan dapat menyebabkan overfitting dan penurunan performa pada data baru. Implementasi sistem ini berhasil mengatasi kelemahan pencatatan manual dengan akurasi prototipe sebesar 71,43%, meskipun keterbatasan kamera masih menjadi kendala. Proyek Akhir ini masih dalam tahap penelitian dan belum diterapkan di industri, sehingga diperlukan pengujian dan pengembangan lebih lanjut sebelum dapat diimplementasikan secara penuh.
=================================================================================================================================
PT. XYZ is committed to implementing the Zero Waste concept in all of its products by processing production waste into liquid fertilizer, which is purchased by vendor partners through a deposit system. However, the manual process of recording vendor tank identities using paper has resulted in paper waste accumulation. To address this issue, the implementation of a vehicle license plate identification system using the Convolutional Neural Network (CNN) method with the YOLOv8 algorithm is proposed. YOLOv8 enables real-time license plate detection with high accuracy, speeding up the recording process and reducing paper waste. Testing results show that the YOLOv8 model with 100 epochs delivers the best performance, achieving a precision of 91,9%, recall of 64,14%, and accuracy of 64,14% in license plate recognition, and precision of 88.68%, recall of 88.73%, and accuracy of 99.93% in character identification. However, the number of epochs and the object's distance from the camera affect system performance, as excessive epochs can lead to overfitting and decreased performance on new data. This system implementation successfully addresses the weaknesses of manual recording with prototype accuracy of 71.43%, although camera limitations remain a challenge. This Final Project is still in the research phase and has not yet been implemented in the industry, thus requiring further testing and development before full implementation.
Item Type: | Thesis (Diploma) |
---|---|
Uncontrolled Keywords: | Identifikasi Plat Nomor, Convolutional Neural Network, YOLOv8, License Plate Identification |
Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK1007 Electric power systems control T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK3070 Automatic control |
Divisions: | Faculty of Vocational > 36304-Automation Electronic Engineering |
Depositing User: | Chata Trista Dyanti |
Date Deposited: | 17 Sep 2024 02:04 |
Last Modified: | 17 Sep 2024 02:08 |
URI: | http://repository.its.ac.id/id/eprint/115643 |
Actions (login required)
View Item |