Drowsiness Detection pada wajah pengemudi dengan kombinasi MCT Adaboost Classifier dengan Histogram Of Oriented Gradients
M. Edu Agritama, Wahyono, S. Kom., Ph.D.
2025 | Tesis | S2 Ilmu Komputer
Untuk
mengidentifikasi kantuk, digunakan metode ekstraksi fitur Histogram of Oriented
Gradients, yang membantu mengenali pola seperti mata terpejam dengna fokus
pada EAR (Eye Aspect Ratio). Metode Histogram of Oriented Gradients digunakan
untuk deteksi kantuk dengan menggabungkannya dengan metode ekstraksi fitur dan
klasifikasi MCT Adaboost Classifier. Hasil ekstraksi fitur dengan Histogram
of Oriented Gradients dan Support Machine Vector menghasilkan nilai akurasi keseluruhan 97%, Kombinasi Histogram of Oriented Gradients dengan Modified Census Transform
dengan klasifikasi Adaboost menghasilkan nilai akurasi keseluruhan 98%, dan evaluasi
PERCLOS pada keseluruhan pengemudi menghasilkan akurasi sebesar 77 %.
In recent times, human activity
has increased significantly, causing many people to prefer using private
vehicles over public vehicles due to efficiency in time and travel. However,
this high level of activity causes many car drivers to experience fatigue and
drowsiness while driving, potentially leading to serious traffic accidents. The
phenomenon of drowsiness or sudden falling asleep while driving is called
microsleep and often occurs in very tired drivers. This problem can be solved
with drowsiness detection. Drowsiness detection in vehicle drivers using facial
recognition. Microsleep, which is falling asleep suddenly for a few seconds,
often occurs in very tired drivers.
To identify drowsiness, the
Histogram of Oriented Gradients feature extraction method is used, which helps
recognize patterns such as closed eyes. The Histogram of Oriented Gradients
method is used for drowsiness detection by combining it with the MCT Adaboost
Classifier feature extraction and classification method. The results of feature
extraction with Histogram of Oriented Gradients and Support Machine Vector
resulted in an overall accuracy value of 97%, the
combination of Histogram of Oriented Gradients with Modified Census Transform
with Adaboost classification resulted in an overall accuracy value of 98%, and the PERCLOS evaluation on the entire driver
resulted in an accuracy of 77%.
Kata Kunci : Drowsiness Detection, Face Recoginition, MCT Adaboost, Histogram Of Oriented Gradients