CD Skripsi

Rancang Bangun Model Deteksi Untuk Tingkat Kematangan Dan Penghitungan Tandan Buah Kelapa Sawit Menggunakan Computer Vision Dengan Algoritma Yolo Dan Bytetrack

---

This study aims to design a system for detecting ripeness levels and automatically counting oil palm Fresh Fruit Bunches (FFB) using a computer vision approach based on YOLOv8 and ByteTrack algorithms. The main issue addressed is the low efficiency and accuracy of the manual sorting process based on ripeness level, which is still commonly used in the palm oil industry. The system is designed to recognize two fruit classes ripe and unripe through RGB value analysis and image data captured using an industrial camera. The object detection model was trained using YOLOv8s, while real-time object tracking and counting were carried out using the ByteTrack algorithm. The dataset consisted of 300 FFB images (15 ripe and 15 unripe samples) and several test videos. Evaluation results showed excellent detection performance, with 100% precision and 99.5% recall for both classes, along with mAP50 of 95.8% and mAP50-95 of 91.3%. Tracking performance yielded MOTA scores of 0.51 for ripe, 0.64 for unripe, and 0.66 for mixed classes, with consistently high IDF1 values across all categories. Furthermore, manual evaluation using video tests showed F1-scores of 0.80 (ripe), 1.00 (unripe), and 0.71 (mixed). These findings demonstrate that the integration of YOLOv8 and ByteTrack provides an effective and accurate solution to support the automation of FFB detection, tracking, and counting processes in the palm oil industry.

Keywords: oil palm, computer vision, YOLOv8, ByteTrack.
.

---

Ketersediaan

Informasi Detail

Judul Seri

-

No. Panggil

2103134782

Penerbit

Pekanbaru : Universitas Riau FMIPA Fisika., 2025

Deskripsi Fisik

-

Bahasa

Indonesia

ISBN/ISSN

-

Klasifikasi

2103134782

Tipe Isi

-

Tipe Media

-

Tipe Pembawa

-

Edisi

-

Subjek

FISIKA

Info Detail Spesifik

-

Pernyataan Tanggungjawab

ERA

Versi lain/terkait

Lampiran Berkas

Komentar

---

Anda harus masuk sebelum memberikan komentar