CD Skripsi

SISTEM PENDETEKSI KELENGKAPAN PEKERJA MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK UNTUK KESELAMATAN BEKERJA

---

Occupational safety in the work environment is a crucial aspect, underlying the
responsibility of every individual in an organization. The fulfillment of work safety
standards has a significant impact on the achievement of the desired targets by an
agency. Conversely, the unavailability of work safety can lead to risks, such as work
accidents, inability to achieve targets, and even unwanted financial losses.
Personal Protective Equipment (PPE) is the main point in maintaining work safety.
Each type of PPE, such as helmets, vests and shoes, has a designation according to
the position of the worker. OHS supervisors directly check the completeness of
workers' PPE in the field, but this manual method can be inefficient and timeconsuming. Therefore, the use of AI technology is a smart solution. The developed
system will detect workers' PPE using a camera. The system will provide a voice to
confirm the completeness of PPE, providing a warning if detected incomplete. AI
technology, especially Machine Learning with a Deep Learning approach, is used
to recognize PPE with the help of a Convolutional Neural Network (CNN). The
application of Computer Vision object detection techniques, such as You Only Look
Once (YOLO), enables real-time object detection. This research uses the Research
and Development (R&D) method which involves a number of systematic steps to
develop a worker equipment detection system based on the Convolutional Neural
Network method using the YOLOv8 algorithm. The detection results are carried out
in two ways, namely image input and in real time. With a high accuracy value, the
system can detect workers' helmets, vests, and shoes with an accuracy of 0.971,
recall of 0.956, and F1 calculation of 96%. This system will later use Arduino to
create an open and closed plank process where if the apd is complete then the plank
will be open and if not the plank is still closed. According to the test, the confidence
of the detection result reaches 90% and at least 60% or more, which shows good
performance. Suggestions for this research are to increase the dataset in the image
and use the latest version of yolo so that the detection results become more accurate
and the results become better.
Keywords: YOLOv8, Convolutional Neural Network, Pengujian Real-time

---

Ketersediaan

Informasi Detail

Judul Seri

-

No. Panggil

2007126282

Penerbit

Pekanbaru : Universitas Riau - F. Teknik - Teknik Elektro., 2024

Deskripsi Fisik

-

Bahasa

Indonesia

ISBN/ISSN

-

Klasifikasi

2007126282

Tipe Isi

-

Tipe Media

-

Tipe Pembawa

-

Edisi

-

Subjek

TEKNIK ELEKTRO

Info Detail Spesifik

-

Pernyataan Tanggungjawab

GUNTUR

Versi lain/terkait

Lampiran Berkas

Komentar

---

Anda harus masuk sebelum memberikan komentar