CD Skripsi

Sistem Optik Pencitraan Multispektral Berbasis Led Dan Metode Deteksi Objek Yolo Untuk Klasifikasi Kematangan Tandan Buah Segar Kelapa Sawit

---

ABSTRACT
The quality of CPO (Crude Palm Oil) can be determined by an effective oil palm FFB (Fresh Fruit Bunch) grading process. Multispectral imaging offers less time image processing, with simple identification system, and low cost. Object detection method is acquired for FFB hence automatic and real-time grading can be achieved. This research aims to develop LED-based multispectral imaging system with object detection method for oil palm FFB ripeness classification. Python-based programs were used in the image acquisition and processing, and object detection model. YOLO (You Only Look Once) algorithm was used as computer vision-based of object detection method. The LED array used as the light source consists of eight LEDs with the wavelengths of: 680, 700, 750, 780, 810, 850, 880, and 900 nm. Oil content and FFA (Free Fatty Acids) were measured by soxhlet extraction and then used to validate the FFB ripeness levels. PCA (Principal Component Analysis) was used for pattern recognition of the relative reflectance intensities of multispectral images for two ripeness categories. Confusion matrix was used to analyze the performance of YOLO model. The research results show that the relative reflectance intensities of ripe FFB are 7.75% higher than unripe FFB in all LED wavelengths. The PCA method was able to differentiate the two ripeness levels of FFB at 97.07% cumulative variance (PC1 = 92.20% and PC2 = 4.87%). The Confusion matrix analysis shows the performance on the YOLO model with : 99.66% for mean Average Precision (mAP), 80% accuracy for ripe FFBs, and 78.57% accuracy for unripe FFBs.
Keyword: Multispectral Imaging, LED, YOLO, Python, Relative reflectance intensity, Oil Content, FFA, PCA, Confusion Matrix

---

Ketersediaan

Informasi Detail

Judul Seri

-

No. Panggil

1903110071

Penerbit

Pekanbaru : Universitas Riau - FMIPA - Fisika., 2023

Deskripsi Fisik

-

Bahasa

Indonesia

ISBN/ISSN

-

Klasifikasi

1903110071

Tipe Isi

-

Tipe Media

-

Tipe Pembawa

-

Edisi

-

Subjek

FISIKA

Info Detail Spesifik

-

Pernyataan Tanggungjawab

TIAR

Versi lain/terkait

Lampiran Berkas

Komentar

---

Anda harus masuk sebelum memberikan komentar