Penerapan Sistem Rekognisi Dan Akuisisi Data Bagcoding Berbasis YOLO11n Dan PaddleOCR Guna Inspeksi Kualitas Cetakan Dan Traceability Produk Pada Lini Pengantongan Pupuk

Putra, Rahadian Dwi Martina (2026) Penerapan Sistem Rekognisi Dan Akuisisi Data Bagcoding Berbasis YOLO11n Dan PaddleOCR Guna Inspeksi Kualitas Cetakan Dan Traceability Produk Pada Lini Pengantongan Pupuk. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221036-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Proses pengantongan pupuk di PT Petrokimia Gresik masih menghadapi permasalahan pada kualitas pencetakan kode produksi (bagcoding) akibat penyumbatan nozzle printer inkjet oleh debu industri serta efek motion blur yang ditimbulkan oleh pergerakan konveyor. Kondisi tersebut menyebabkan sebagian bagcoding tidak dapat terbaca dengan baik sehingga inspeksi visual yang masih dilakukan secara manual berpotensi meloloskan produk cacat dan belum mampu mendukung ketertelusuran (traceability) secara digital. Penelitian ini bertujuan mengimplementasikan sistem rekognisi dan akuisisi data bagcoding berbasis deep learning yang mampu melakukan inspeksi kualitas sekaligus mengintegrasikan hasil pembacaan ke dalam Smart Bagging Ecosystem. Sistem yang akan diimplementasikan menerapkan arsitektur dua tahap (two-stage pipeline), yaitu YOLO11n sebagai pendeteksi lokasi bagcoding dan PaddleOCR sebagai mesin rekognisi karakter. Untuk meningkatkan kualitas citra sebelum proses rekognisi, diterapkan tahapan preprocessing berupa deblurring dan dilasi morfologi. Hasil pembacaan selanjutnya dikirimkan melalui protokol MQTT menuju mikrokontroler ESP8266 untuk mengendalikan Early Warning System (EWS), sedangkan data produk yang memenuhi kriteria disimpan ke dalam basis data sebagai identitas digital (digital birth certificate) produk. Hasil pengujian menunjukkan bahwa model YOLO11n mampu mendeteksi area bagcoding dengan akurasi tinggi, ditunjukkan oleh nilai mean Average Precision sebesar 98,02% serta waktu inferensi rata-rata 11,39 ms sehingga memenuhi kebutuhan deteksi real-time. Pada tahap rekognisi, PaddleOCR menghasilkan Exact Match Accuracy sebesar 83,33% dan Near-Match Accuracy sebesar 97,67%, dengan nilai Character Error Rate (CER) sebesar 1,37%. Pengujian komunikasi MQTT juga menunjukkan tingkat keandalan yang tinggi dengan Packet Delivery Ratio sebesar 99,8%, sedangkan keseluruhan sistem memiliki latensi kurang dari 3 detik. Berdasarkan hasil tersebut, sistem yang dikembangkan dinyatakan mampu mendukung inspeksi kualitas bagcoding dan akuisisi data produk secara real-time pada lini pengantongan pupuk.
========================================================================================================================
The fertilizer bagging process at PT Petrokimia Gresik continues to face challenges related to the readability of production codes (bagcoding), primarily caused by inkjet printer nozzle clogging due to industrial dust and motion blur induced by conveyor movement. These conditions often result in unreadable bagcoding, making manual visual inspection prone to passing defective products while failing to support digital product traceability. This study aims to implement a deep learning-based bagcoding recognition and data acquisition system capable of performing quality inspection while integrating the inspection results into the Smart Bagging Ecosystem. The proposed system adopts a two-stage pipeline consisting of YOLO11n for bagcoding localization and PaddleOCR for optical character recognition. To improve image quality prior to text recognition, image preprocessing techniques, including deblurring and morphological dilation, were applied. The recognition results were subsequently transmitted through the MQTT protocol to an ESP8266 microcontroller to control the Early Warning System (EWS), while products meeting the inspection criteria were automatically recorded in a database as their digital birth certificates. Experimental results demonstrated that YOLO11n achieved high detection performance with a mean Average Precision of 98.02% and an average inference time of 11.39 ms, satisfying real-time detection requirements. In the recognition stage, PaddleOCR achieved an Exact Match Accuracy of 83.33%, a Near-Match Accuracy of 97.67%, and a Character Error Rate (CER) of 1.37%. MQTT communication also demonstrated high reliability with a Packet Delivery Ratio (PDR) of 99.8%, while the overall end-to-end system latency remained below 3 seconds. These results indicate that the proposed system is capable of supporting real-time bagcoding quality inspection and automated product data acquisition in fertilizer bagging production lines.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Bagcoding, Inspeksi Otomatis, MQTT, PaddleOCR, Traceability, YOLO11n, Automatic Inspection, Bagcoding, MQTT, PaddleOCR, Traceability, YOLO11n.
Subjects:	H Social Sciences > HD Industries. Land use. Labor > HD3656 Inspection. Factory inspection Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA336 Artificial Intelligence T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing. T Technology > TS Manufactures > TS156 Quality Control. QFD. Taguchi methods (Quality control)
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Rahadian Dwi Martina Putra
Date Deposited:	05 Aug 2026 09:05
Last Modified:	05 Aug 2026 09:05
URI:	http://repository.its.ac.id/id/eprint/144082

Actions (login required)

View Item