Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/150463
Title: Deteksi Kerusakan Mekanis Buah Tomat Ceri Berdasarkan Citra Fluoresen dengan Metode Convolutional Neural Network
Other Titles: Detection of Mechanical Damage in Cherry Tomatoes Based on Fluorescent Imagery with Convolutional Neural Network
Authors: Solahudin, Mohamad
Widodo, Slamet
Innocensia, Tersia Ralesmanti
Issue Date: 2024
Publisher: IPB University
Abstract: Tomat ceri merupakan komoditas buah dengan nilai ekonomis tinggi bagi Indonesia. Namun, hingga saat ini, kerusakan mekanis sering terjadi dalam berbagai tahap, mulai dari panen hingga penyimpanan. Deteksi kerusakan mekanis pada tomat ceri umumnya masih dilakukan secara manual, yakni dengan menggunakan mata secara langsung. Proses tersebut rentan terhadap subjektivitas, kelelahan, dan kesalahan pandang. Tujuan penelitian ini mengembangkan suatu metode deteksi kerusakan mekanis buah tomat ceri seperti memar dan sayat dengan citra fluoresen menggunakan metode convolutional neural network (CNN). Hasil penelitian diharapkan mampu memberikan kemudahan dalam deteksi kerusakan mekanis pada buah tomat ceri secara lebih akurat yang nantinya dapat diimplementasikan untuk proses sortasi. Penelitian ini menggunakan lampu ultraviolet tipe-c, longpass filter, circular polarizer filter. Penggunaan lampu ultraviolet dan filter-filter kamera tersebut untuk menghasilkan citra fluoresen sesuai dengan karakteristik buah tomat ceri. Penelitian ini menggunakan algoritma YOLOv8 pada Google Colaboratory. Penelitian ini terdiri dari beberapa tahapan: (1) pengumpulan dataset, (2) identifikasi dan pelabelan data, (3) training dataset, (4) pengujian model deep learning, (5) analisis hasil uji fungsional model deep learning. Nilai akurasi, presisi, recall, dan F1-score berturut-turut sebesar 95,33%; 100,00%; 87,78%, 93,49%. Hasil tersebut menunjukkan bahwa model yang dibangun mampu mendeteksi kerusakan mekanis buah tomat ceri dengan akurat dan presisi.
Cherry tomatoes are an economically important crop for Indonesia. However, until now, mechanical damage has often accured at various stages, from harvest to storage. Damage detection is usually done manually through visual inspection. This process is susceptible to subjective, fatigue, and errors of view. The research aims to develop a damage detection method for cherry tomatoes using fluorescence imaging and a convolutional neural network (CNN) approach. The research results are expected to easier detect mechanical damage to cherry tomatoes more accurately, which can later be implemented in the sorting process. This research uses a type-c ultraviolet lamp, a longpass filter, a circular polarizer filter. The use of ultraviolet lights and camera filters to produce fluorescent images according to the characteristics of cherry tomatoes. This research uses the YOLOv8 algorithm on Google Colaboratory. This research consists of several stages: (1) dataset collection, (2) data identification and labeling, (3) training dataset, (4) deep learning model testing, and (5) analysis of functional test results of the deep learning model. The accuracy, precision, recall, and F1-score values were 95.33%; 100.00%; 87.78%, 93.49%. These results show that the model built can detect mechanical damage to cherry tomatoes with accuracy and precision.
URI: http://repository.ipb.ac.id/handle/123456789/150463
Appears in Collections:UT - Agricultural and Biosystem Engineering

Files in This Item:
File Description SizeFormat 
Cover (1).pdfCover444.42 kBAdobe PDFView/Open
F1401201099_Tersia Ralesmanti Innocensia.pdfFull Teks1.92 MBAdobe PDFView/Open
Lampiran.pdfLampiran235.62 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.