CD Tesis

Prediksi Tingkat Kematangan Tandan Buah Segar Kelapa Sawit Dan Kandungan Kimia Menggunakan Pencitraan Multispektral Dan Algoritma Yolo

---

SUMMARY
Nisa Arpyanti, NIM. 2210246740. Prediction of Ripeness Level and Chemical
Content of Oil Palm Fresh Fruit Bunches Usimh Multispectral Imaging and
YOLO algorithm, supervised by Prof. Dr. Minarni, M.Sc and Dr.-Ing Rahmondia
Nanda Setiadi, M.Si.
The main issue addressed in this study is the limitation of the current sorting
system for oil palm Fresh Fruit Bunches (FFB), which is still performed manually
(using labor) and traditionally (based on human visual assessment), leading to a
lack of consistency and objectivity in determining ripeness levels. RGB imaging
methods have previously been used for ripeness detection, but they operate only
within the visible spectrum, capturing three main wavelengths. In contrast,
multispectral imaging offers a broader spectral range from visible to near-infrared
allowing access to visual and chemical features not detectable by RGB cameras,
such as fruit skin pigments and chemical compounds like oil content and Free Fatty
Acids (FFA).
This study utilized a multispectral imaging system previously developed in
earlier research, consisting of a monochrome CMOS camera and eight LEDs with
wavelengths ranging from 680 to 900 nm. The system was used to capture images
of FFB in two ripeness categories: unripe and ripe. The resulting reflectance
intensity data were analyzed using Principal Component Analysis (PCA) to classify
the data based on three approaches: (1) reflectance intensity according to ripeness,
(2) oil content according to ripeness, and (3) FFA content according to ripeness.
Laboratory testing of the chemical content was conducted to examine the
relationship between fruit ripeness and its chemical parameters.
The results showed that PCA was effective in clustering reflectance intensity
data between unripe and ripe fruits. In the analysis based on FFA content, the first
principal component (PC1) accounted for 91,2% of the variance, while the second
component (PC2) explained 6,21%. For oil content, PC1 accounted for 91,2% of
the variance, and PC2 for 6,21%. The YOLOv8 object detection model
demonstrated strong performance in ripeness classification, achieving 93.3%
accuracy, 100% precision, 100% recall, and a 100% F1-score. These findings
indicate that the combination of multispectral imaging, PCA analysis, and the
YOLO algorithm can serve as a non-destructive and efficient method for
automatically classifying the ripeness and chemical composition of oil palm FFB.
This method has the potential to be applied in the automated sorting process within
the palm oil industry to support improved quality and efficiency in Crude Palm Oil
(CPO) production.
Key words: Multispectral imaging, Oil palm fresh fruit bunch, Oil content, Free
fatty acid, YOLOv8

---

Ketersediaan

Informasi Detail

Judul Seri

-

No. Panggil

2210246740

Penerbit

Pekanbaru. : Universitas Riau – PASCA– Magister Fisika., 2025

Deskripsi Fisik

-

Bahasa

Indonesia

ISBN/ISSN

-

Klasifikasi

2210246740

Tipe Isi

-

Tipe Media

-

Tipe Pembawa

-

Edisi

-

Subjek

Magister Fisika

Info Detail Spesifik

-

Pernyataan Tanggungjawab

RIDHO

Versi lain/terkait

Lampiran Berkas

Komentar

---

Anda harus masuk sebelum memberikan komentar