Search for collections on Undip Repository

SISTEM DETEKSI KERUSAKAN KERAMIK TABLEWARE MENGGUNAKAN MORPHOLOGICAL ENHANCEMENT DAN CONVOLUTIONAL NEURAL NETWORK

RAHMAYUNA, Novita and Adi, Kusworo and Kusumaningrum, Retno (2022) SISTEM DETEKSI KERUSAKAN KERAMIK TABLEWARE MENGGUNAKAN MORPHOLOGICAL ENHANCEMENT DAN CONVOLUTIONAL NEURAL NETWORK. Masters thesis, UNIVERSITAS DIPONEGORO.

[img] Text
COVER.pdf

Download (925kB)
[img] Text
BAB I.pdf

Download (152kB)
[img] Text
BAB II.pdf

Download (441kB)
[img] Text
BAB III.pdf
Restricted to Repository staff only

Download (595kB)
[img] Text
BAB IV.pdf
Restricted to Repository staff only

Download (589kB)
[img] Text
BAB V.pdf

Download (75kB)
[img] Text
DAFTAR PUSTAKA.pdf

Download (150kB)
[img] Text
DAFTAR LAMPIRAN.pdf
Restricted to Repository staff only

Download (103kB)

Abstract

Penelitian ini dilatarbelakangi dengan adanya kesalahan yang dilakukan dalam proses pemeriksaan fisik pada keramik tableware. Hal ini disebabkan karena faktor human-error yang dilakukan oleh manusia. Salah satu cara untuk menangani permasalahan tersebut yaitu dengan memanfaatkan teknologi komputer untuk mengenali kerusakan pada keramik tableware. Pengolahan citra digital dapat diterapkan untuk menangani permasalahan ini. Objek penelitian diambil langsung di PT. Sango Internasional Semarang dengan kriteria keramik tableware dalam bentuk piring dan berwarna putih. Penelitian ini mengusulkan metode morphological enhancement dan Convolutional Neural Network (CNN) untuk mendeteksi letak dan jenis kerusakan yang ada pada keramik tableware. Data sejumlah 56 video dengan empat jenis kondisi kerusakan digunakan pada penelitian ini. Kinerja model yang dihasilkan diukur menggunakan akurasi klasifikasi. Proses pelatihan dilakukan dengan menggunakan kombinasi parameter CNN yaitu solver dengan nilai SGDM dan ADAM, mini batch size dengan nilai 32, 64 dan 128, serta learning rate dengan nilai 0.0001, 0.00001, dan 0.000001. Model terbaik didapatkan dengan parameter solver ADAM, parameter mini batch size dengan ukuran 32 dan parameter learning rate dengan nilai 0.00001. Model ini menghasilkan akurasi sistem sebesar 99,82%. Selain itu, letak kerusakan ditandai dengan bounding box pada kerusakan yang ada pada keramik tableware dan tingkat kerusakan sistem dihitung berdasarkan luasan area keramik tableware.
Kata Kunci : Convolutional Neural Network, deep learning, morphological enhancement, deteksi kerusakan

This research is motivated by the existence of errors made in the physical examination process on ceramic tableware. This is due to the human-error factor committed by humans. One way to deal with this problem is to use computer technology to identify damage to ceramic tableware. Digital image processing can be applied to deal with this problem. The object of research is taken directly at PT. Sango Internasional Semarang with the criteria of ceramic tableware in the form of plates and white. This study proposes morphological enhancement and Convolutional Neural Network (CNN) methods to detect the location and type of damage on ceramic tableware. A total of 56 videos with four types of damage conditions were used in this study. The performance of the resulting model is measured using classification accuracy. The training process is carried out using a combination of CNN parameters, namely solver with values of SGDM and ADAM, mini batch size with values of 32, 64 and 128, and learning rate with values of 0.0001, 0.00001, and 0.00001. The best model was obtained with the ADAM solver parameter, the mini batch size parameter with a size of 32 and the learning rate parameter with a value of 0.00001. This model produces a system accuracy of 99,82%. In addition, the location of the damage is indicated by a bounding box on the existing damage to the ceramic tableware and the level of system damage is calculated based on the area of the ceramic tableware.
Keyword : Convolutional Neural Network, deep learning, morphological enhancement, defect detection

Item Type: Thesis (Masters)
Uncontrolled Keywords: Convolutional Neural Network, deep learning, morphological enhancement, deteksi kerusakan
Subjects: Sciences and Mathemathic
Divisions: Postgraduate Program > Master Program in Information System
Depositing User: ekana listianawati
Date Deposited: 20 Feb 2023 03:54
Last Modified: 20 Feb 2023 03:54
URI: https://eprints2.undip.ac.id/id/eprint/11772

Actions (login required)

View Item View Item