PENGEMBANGAN SISTEM COMPUTER VISION UNTUK DETEKSI PARAMETER MUTU KADAR AFLATOKSIN B1 PADA BIJI KOPI MENGGUNAKAN METODE PENCITRAAN FLUORESENSI ULTRAVIOLET

• Main Author: Adlani, Arsyi
• Format: Final Project
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/77379
• Institution: Institut Teknologi Bandung

Coffee beans are one of the potential national export commodities in the non-oil and gas sector. Throughout the year 2021, the recorded total export volume of Indonesian coffee beans reached 380,17 thousand tons with a total export value amounting to US$ 842,52 million. These figures have positioned Indonesia as the fourth-largest producer of coffee beans in the world. To ensure the quality of the traded national coffee beans, the government regulates the quality requirements for coffee beans through the Indonesian National Standard (SNI) 01-2907-2008 on Coffee Beans. One of the stipulated requirements is that coffee beans must be free from filamentous fungi (mold). However, the current process of testing the quality parameters of coffee beans is still conducted manually by testers, resulting in a lengthy and inefficient process. The methods currently employed for detecting levels of aflatoxin B1 typically rely on chromatography-based techniques, including thin-layer chromatography (TLC), high-performance liquid chromatography (HPLC), and enzyme-linked immunosorbent assay (ELISA). Nevertheless, these methods are generally destructive, time-consuming, and expensive due to the need for specialized equipment for the purification process. Addressing these challenges, fluorescence imaging has been developed as an alternative method for aflatoxin B1 detection in food commodities. The development of ultraviolet fluorescence imaging integrated with a computer vision system for image processing and coffee bean quality quantification could mark the initial step towards supporting reliable quality assurance efforts, enhancing food safety, and elevating the competitiveness of agricultural products, particularly coffee beans, as a major national export commodity. A computer vision system based on the Random Forest classical classification model, combined with region of interest (ROI) extraction and feature selection is developed and yields the best performance, achieving an accuracy rate of 89,87% and an average precision rate of 89,68% for classifying aflatoxin B1 levels in UV fluorescence images of coffee beans.