Design Of Computer Vision Technology On Loading Ramp, For Effective FFB Categorization To Increase Process Optimization In Palm Oil Mills

Kamil, Muhammad Kalif Qisthi (2024) Design Of Computer Vision Technology On Loading Ramp, For Effective FFB Categorization To Increase Process Optimization In Palm Oil Mills. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5010201115-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (5MB) | Request a copy

Abstract

Indonesia, sebagai pemilik lahan kelapa sawit terluas dan juga produsen terbesar Crude Palm Oil (CPO) dan Plam Kernel Oil (PKO) beserta produk sampingannya, merupakan oportunitas keunggulan yang perlu ditetapkan. Hal ini mendorong industri kelapa sawit untuk melakukan perbaikan terus menerus improvisasi. Pada industri pengolahan kelapa sawit, serangkaian proses diperlukan dimulai dari sortir tandan buah segar (TBS) hingga perebusan, penggilingan, klarifikasi, sampai dihasilkan CPO, dan PKO. Proses inspeksi dan pemilahan TBS yang dilakukan pada area penerimaan Loading Ramp menjadi salah satu proses yang sangat krusial. Sebab rata-rata aktivitas inspeksi dan pemilahan TBS di area Loading Ramp saat ini dilakukan secara manual dan dengan single-visual, rawan terjadinya ketidakakuratan dan ketidakkonsistenan pemilahan jenis kategori TBS, TBS yang tidak matang akan mempengaruhi volume dan kualitas CPO yang dihasilkan. Penelitian ini mengusulkan aplikasi teknologi Computer Vision, merupakan teknologi untuk penerjemahan media visual, teknologi ini digunakan untuk melakukan proses optimasi inspeksi dan klasifikasi TBS di area Loading Ramp. Untuk meningkatkan praktik pemeriksaan yang saat ini diterapkan, merancang teknologi software Computer Vision dilakukan untuk meningkatkan kegiatan pemeriksaan dan pemilahan TBS. Identifikasi kematangan melalui warna dan Brondolan diterapkan untuk menjadi akumulasi dan klasifikasi kematangan TBS karena kedua karakteristik merupakan bukti visual kematangan yang sangat menonjol tepat untuk penggunaan Computer Vision, menggunakan methode Deep Learning (CNN) dan Machine Learning (HOG-SVM) berturut-turut. Tahap pelatihan mencapai akurasi sebesar 86,69% untuk CNN dan 62,83% untuk HOG-SVM. Pada tahap pengujian diperoleh akurasi penyortiran untuk buah terlalu matang sebesar 93%, buah matang sebesar 86%, dan buah kurang matang sebesar 92%, dengan akurasi pengambilan akurasi secara keseluruhan sebesar 90% dan membutuhkan waktu kurang lebih 30 detik.
=================================================================================================================================
Indonesia, as the owner of the largest oil palm plantations and the largest producer of Crude Palm Oil (CPO) and Palm Kernel Oil (PKO) along with their by-products, represents an opportunity for establishing a competitive advantage. This drives the palm oil industry to continually improve and innovate. In the palm oil processing industry, a series of processes are required, starting from sorting fresh fruit bunches (FFB) to boiling, milling, clarification, until CPO and PKO are produced. The inspection and sorting process of FFB conducted in the Loading Ramp reception area is one of the most crucial processes. This is because the average FFB inspection and sorting activity in the Loading Ramp area is currently done manually and with single-visual inspection, making it prone to inaccuracies and inconsistencies in sorting FFB categories. Unripe FFB will affect the volume and quality of the resulting CPO. This research proposes the application of Computer Vision technology, a technology for interpreting visual media, to optimize the inspection and classification process of FFB in the Loading Ramp area. To improve the current inspection practices, the design of Computer Vision software technology aims to enhance FFB inspection and sorting activities. Maturity identification through color and loose fruits is applied to accumulate and classify FFB maturity, as these two characteristics are highly prominent visual evidence of maturity, suitable for the use of Computer Vision using Deep Learning (CNN) and Machine Learning (HOG-SVM) methods respectively. The training stage achieved accuracies of 86.69% for CNN and 62.83% for HOG-SVM. In the testing stage, sorting accuracies were obtained for overripe fruit at 93%, ripe fruit at 86%, and underripe fruit at 92%, with an overall accuracy for decision-making at 90% and took time of approximately 30 seconds.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Fresh Fruit Bunches, Computer Vision, Agriculture, Palm Oil, Tandan Buah Segar, Computer Vision, Pertanian, Kelapa Sawit.
Subjects:	T Technology > TS Manufactures > TS155 Production control. Production planning. Production management T Technology > TS Manufactures > TS156 Quality Control. QFD. Taguchi methods (Quality control) T Technology > TS Manufactures > TS176 Manufacturing engineering. Process engineering (Including manufacturing planning, production planning) T Technology > TS Manufactures > TS183 Manufacturing processes. Lean manufacturing.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Kalif Qisthi Kamil
Date Deposited:	18 Jul 2024 07:32
Last Modified:	18 Jul 2024 07:32
URI:	http://repository.its.ac.id/id/eprint/108446

Actions (login required)

View Item