Putra, Yesaya Kharismael Kurniawan (2022) Klasifikasi Skala Kantuk Karolinska Berdasarkan Nilai Bukaan Mata (Perclos) Menggunakan Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.
|
Text
07211840000021-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (3MB) | Request a copy |
Abstract
Mengantuk merupakan reaksi alamiah tubuh manusia yang terkadang dapat mengganggu aktivitas dari manusia itu sendiri, termasuk saat mengemudi. Pendeteksian kantuk pengemudi dibutuhkan untuk memperkecil kemungkinan terjadinya kecelakaan. Saat ini, pendeteksian kantuk mempunyai dua metode, yaitu metode intrusive dan non-intrusive. Pendeteksian kantuk dengan metode intrusive atau kontak fisik, pada umumnya menggunakan EEG. Metode intrusive dinilai tidak praktis dan cukup mengganggu pengemudi karena membutuhkan alat berupa sensor yang ditempelkan pada beberapa bagian tubuh. Oleh karena itu, dibutuhkan metode non-intrusive untuk melakukan pendeteksian yang menggunakan metode deep learning atau visi komputer. Pendeteksian non-intrusive mempunyai berbagai cara, tetapi diantara metode-metode yang ada, metode dengan penggunaan terbanyak adalah berdasarkan nilai bukaan mata (PERCLOS). Hasil yang didapatkan dari penelitian ini adalah model dapat mengklasifikasikan skala kantuk sesuai Skala Kantuk Karolinska (SKK) berdasarkan nilai bukaan mata (PERCLOS) dengan arsitektur terbaik menggunakan masukan (input) dari data hasil oversampling yang memiliki akurasi sebesar 86% dan 89%.
==============================================================================================================================
Drowsiness is a natural reaction of the human body that can sometimes interfere with the activities of humans themselves, including when driving. Detection of driver drowsiness is needed to minimize the possibility of accidents. Currently, there are two methods of detecting sleepiness, namely intrusive and non-intrusive methods. Detection of drowsiness by intrusive methods or physical contact, generally using EEG. The intrusive method is considered impractical and quite disturbing to the driver because it requires a device in the form of sensors attached to several parts of the body. Therefore, a non-intrusive method is needed to detect using deep learning or computer vision methods. Non-intrusive detection has various methods, but among the existing methods, the method with the most use is based on the eye-opening value (PERCLOS). The results obtained from this study are that the model can classify the sleepiness scale according to the Karolinska Sleepiness Scale (SKK) based on the eye-opening value (PERCLOS) with the best architecture using input (input) from the result data oversampling which has an accuracy of 86% and 89%. Keywords: Drowsiness, Human, Eyes, Deep Learning.
| Item Type: | Thesis (Other) |
|---|---|
| Additional Information: | RSKom 006.42 Put k-1 2022 |
| Uncontrolled Keywords: | Kantuk. Manusia. Mata. Deep Learning. Drowsiness. Human. Eyes. Deep Learning. |
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science. EDP |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis |
| Depositing User: | Mr. Marsudiyana - |
| Date Deposited: | 17 Jun 2026 01:24 |
| Last Modified: | 17 Jun 2026 01:24 |
| URI: | http://repository.its.ac.id/id/eprint/133830 |
Actions (login required)
![]() |
View Item |
