Alhakim, Rahmat Afidan (2024) Sistem Inspeksi Visual Untuk Kondisi Box Dan Sticker Produk Coil Nyamuk Dengan Menggunakan Metode Convolutional Neural Network (CNN). Diploma thesis, Institut Teknologi Sepuluh Nopember.
Text
2040201133_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (5MB) | Request a copy |
Abstract
Pada perusahaan manufaktur merupakan perusahaan yang membuat bahan mentah menjadi produk jadi. Sebelum didistribusikan produk jadi akan dikemas. Produk yang akan dipakai dalam proyek akhir ini ialah coil nyamuk. Coil nyamuk yang telah di produksi akan dikemas dalam box dan diberi label barcode pada kemasannya. Pada produk yang didistribusikan ke luar negeri, pada pengemasan terdapat pemberian sticker label barcode ekstra yang masih dipasang secara manual. Dari pengambilan data lapangan, error dalam pemasangan manual terjadi sejumlah 24 dari 240 produk dalam posisi label. Selain penempatan label terdapat kecacatan kemasan produk berupa penyok sejumlah 11. Dalam pemasangan dan pengecekan label memakan waktu 45 menit dengan waktu pemasangan dan pengecekan yang tidak rata dengan waktu paling cepat 2 detik dan waktu paling lama 4 menit dengan rata pemasangan dan pengecekan 10 detik. Dalam penelitian ini, dibuat pengembangan sebuah sistem labelling otomatis yang menggunakan You Only Look Once (YOLO) sebagai inspeksi masalah penempatan label dan pengecekan kualitas kemasan produk. Sistem ini menggunakan raspberry pi dan motor stepper untuk pemasangan label dan konveyor. Selain itu, digunakan kamera dengan raspberry pi sebagai mini PC untuk memproses data visual dan mengontrol proses penyortiran secara otomatis. Digunakan pendekatan YOLO sebagai metode deteksi objek, yang memungkinkan pengenalan objek secara real-time dan meratakan waktu dalam inspeksi visual memungkinkan sistem untuk secara akurat mendeteksi keadaan kemasan serta posisi penempatan label dengan tingkat keakuratan yang tinggi. Penelitian ini diharapkan dapat memberikan kontribusi positif terhadap hasil dan kualitas produksi, serta mengurangi potensi kesalahan manusia dalam proses penempatan label dan pada produk yang didistribusikan ke luar negeri. Dari hasil test yang dilakukan untuk YoloV5m memiliki F1 score 73.43%, Presisi 73.43%, Recall 73.43%, dan akurasi 79.76%. YoloV5s memiliki F1 score 72.86%, Presisi 72.3%, Recall 73.43%, dan akurasi 79.16%. Menunjukkan bahwa YoloV5m memiliki nilai akurasi tertinggi.
=================================================================================================================================
A manufacturing company is a company that makes raw materials into finished products. Before being distributed the finished product will be packaged. The product that will be used in this final project is a mosquito coil. Mosquito coils that have been produced will be packaged in boxes and labeled with barcodes on the packaging. In products that are distributed overseas, the packaging has extra barcode label stickers that are still installed manually. From field data collection, errors in manual installation occurred in 24 out of 240 products in the label position. In addition to label placement, there are product packaging defects in the form of dents totaling 11. In installing and checking labels, it takes 45 minutes with uneven installation and checking times with the fastest time of 2 seconds and the longest time of 4 minutes with an average installation and checking of 10 seconds. In this research, the development of an automatic labeling system that uses You Only Look Once (YOLO) as an inspection of label placement problems and checking the quality of product packaging is made. This system uses raspberry pi and stepper motor for label installation and conveyor. In addition, a camera is used with a raspberry pi as a mini PC to process visual data and control the sorting process automatically. The YOLO approach is used as an object detection method, which enables real-time object recognition and time leveling in visual inspection allowing the system to accurately detect the state of the packaging as well as the label placement position with a high degree of accuracy. This research is expected to make a positive contribution to production output and quality, as well as reduce the potential for human error in the label placement process and in products distributed overseas. From the test results conducted for YoloV5m, it has an F1 score of 73.43%, precision of 73.43%, recall of 73.43%, and accuracy of 79.76%. YoloV5s has an F1 score of 72.86%, Precision 72.3%, Recall 73.43%, and accuracy 79.16%. Shows that YoloV5m has the highest accuracy value.
Item Type: | Thesis (Diploma) |
---|---|
Uncontrolled Keywords: | You Only Look Once (YOLO), Inspeksi Visual, Convolutional Neural Network (CNN), Visual Inspection |
Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK3070 Automatic control T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7888.3 Digital computers |
Divisions: | Faculty of Vocational > 36304-Automation Electronic Engineering |
Depositing User: | Rahmat Afidan Alhakim |
Date Deposited: | 20 Sep 2024 04:51 |
Last Modified: | 20 Sep 2024 04:51 |
URI: | http://repository.its.ac.id/id/eprint/115662 |
Actions (login required)
View Item |