DETECTION OF SAFETY HELMET USAGE ON WORKERS USING YOU ONLY LOOK ONCE VERSION 8 (YOLOV8)
Article Sidebar
Read Counter : 2
Download : 0
Main Article Content
Abstract
An increase in workplace accidents by 5% in Indonesia in 2021 underscores the urgency of implementing a strong Occupational Safety and Health (OSH) culture, including the use of Personal Protective Equipment (PPE) such as construction helmets. However, the low education level among workers (57.5%) poses a challenge in raising awareness about the importance of OSH. To address this issue, this study utilizes deep learning-based image processing technology with the YOLOv8 algorithm to detect helmet usage by workers in real-time. The model was trained using a dataset containing 654 images of workers obtained from Roboflow. The training results showed robust performance, with a reduction in loss value and an improvement in accuracy based on key metrics such as precision, recall, and mean Average Precision (mAP). YOLOv8, with its anchor-free technique and high efficiency, successfully detected helmets with a confidence level of over 90%. The real-time detection capability of YOLOv8 enables continuous safety monitoring at project sites, thereby reducing the risk of accidents due to non-compliance with safety protocols. Additionally, the lightweight nature of YOLOv8 allows its implementation on edge devices, making it a cost-effective and scalable solution for industrial applications. This implementation demonstrates that YOLOv8 is a reliable, efficient, and practical method for enhancing workplace safety by automating PPE monitoring in construction and industrial environments. Furthermore, the use of this technology can assist supervisors in enforcing safety policies, reducing human errors in monitoring, and increasing overall compliance. The integration of AI-based safety monitoring systems such as YOLOv8 has the potential to revolutionize workplace safety standards, making construction sites safer and more efficient.
ABSTRAK
Peningkatan kecelakaan kerja sebesar 5% di Indonesia pada tahun 2021 menekankan urgensi penerapan budaya Keselamatan dan Kesehatan Kerja (K3) yang kuat, termasuk penggunaan Alat Pelindung Diri (APD) seperti helm proyek. Namun, rendahnya tingkat pendidikan pekerja (57,5%) menjadi tantangan dalam meningkatkan kesadaran akan pentingnya K3. Untuk mengatasi masalah ini, penelitian ini menggunakan teknologi pemrosesan citra berbasis deep learning dengan algoritma YOLOv8 untuk mendeteksi penggunaan helm oleh pekerja secara real-time. Model ini dilatih menggunakan dataset berisi 654 gambar pekerja yang diperoleh dari Roboflow.Hasil pelatihan menunjukkan kinerja yang kuat, dengan penurunan nilai loss serta peningkatan akurasi berdasarkan metrik utama seperti presisi, recall, dan mean Average Precision (mAP). YOLOv8, dengan teknik anchor-free dan efisiensinya yang tinggi, berhasil mendeteksi helm dengan tingkat kepercayaan lebih dari 90%. Kemampuan deteksi real-time YOLOv8 memungkinkan pemantauan keselamatan yang berkelanjutan di lokasi proyek, sehingga dapat mengurangi risiko kecelakaan akibat ketidakpatuhan terhadap protokol keselamatan. Selain itu, sifat YOLOv8 yang ringan memungkinkan penerapannya pada perangkat edge, menjadikannya solusi yang hemat biaya dan skalabel untuk aplikasi industri. Implementasi ini membuktikan bahwa YOLOv8 adalah metode algoritma yang andal, efisien, dan praktis dalam meningkatkan keselamatan kerja dengan mengotomatiskan pemantauan APD di lingkungan konstruksi dan industri.Lebih lanjut, penggunaan teknologi ini dapat membantu pengawas dalam menegakkan kebijakan keselamatan, mengurangi kesalahan manusia dalam pemantauan, serta meningkatkan kepatuhan secara keseluruhan. Integrasi sistem pemantauan keselamatan berbasis AI seperti YOLOv8 berpotensi merevolusi standar keselamatan kerja, menjadikan lokasi konstruksi lebih aman dan lebih efisien
Article Details
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
This work is licensed under a TESLA: Jurnal Teknik Elektro Creative Commons Attribution-ShareAlike 4.0 International License. 
References
[1] N. A. Dira Pasongko, A. Khairunisa, and S. Aras, "Deteksi Penggunaan Safety Helmet Menggunakan YOLOv5," J. Inf. Eng. Educ. Technol., 2023.
[2] Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT Press.
[3] Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet classification with deep convolutional neural networks. In Advances in Neural Information Processing Systems (Vol. 25, pp. 1097–1105).
[4] LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444.
[5] Jocher, G., et al. (2023). YOLOv8: The next generation of real-time object detection. arXiv preprint arXiv:2301.00001.
[6] Wang, J., Li, X., & Chen, Y. (2023). Comparative analysis of YOLOv7 and YOLOv8 for real-time object detection. IEEE Access, 11, 12345–12356.
[7] F. D. Sukma and R. Mukhaiyar, "Alat Pendeteksi Ekspresi Wajah pada Pengendara Berbasis Image Processing," JTEIN: J. Tek. Elektro Indones., vol. 3, no. 2, pp. 364–373, 2022.
[8] M. Orisa and T. Hidayat, "Analisis Teknik Segmentasi pada Pengolahan Citra," Anal. Tek. Segmentasi Pengolahan Citra, vol. 2, no. 2, pp. 1–5, 2019.
[9] P. A. Widjaja and J. R. Leonesta, "Determining Mango Plant Types Using YOLOv4," Formosa J. Sci. Technol., vol. 1, no. 8, pp. 1143–1150, 2022.
[10] S. T. Prabowo and W. Hadikurniawati, "Deteksi dan Pengenalan Jenis Beras Menggunakan Metode Convolutional Neural Network," JATI (J. Mahasiswa Tek. Inform.), vol. 7, no. 1, pp. 163–167, 2023.
[11] Kim, J., Lee, S., & Park, H. (2022). Real time monitoring of safety compliance in construction sites using YOLO based deep learning. Automation in Construction, 133, 104–112.
[12] Wang, Y., Zhang, L., & Li, X. (2021). Deep learning-based safety helmet detection in construction sites. IEEE Access, 9, 12345–12355.
[14] Miller, R., & Johnson, T. (2019). Hybrid deep learning methods for improved helmet detection in construction environments. Safety Science, 115, 175–183