Abstract:
This engineering project objective to develop an OpenCV application for image processing and real-time eye detection using a camera to monitor drivers' drowsiness. The project uses a sample group of 10 students' faces from the Faculty of Engineering, Srinakharinwirot University. The program for detecting facial features and eye blinking is written in Python. More than 18 blinks/min is considered normal, 12-18 blinks/min is considered level 1 drowsiness, less than 12 blinks/min is considered level 2 drowsiness, and if you close your eye for 2 sec is considered level 3 drowsiness. From the test results in the first three levels, the second root mean square error (RMSE) was used as an indicator of eye blink behavior. The sample group had RMSE values that conform to the standard in the range of 0 - 1 , consisting of 0 , 6 , and 9 in sequent. And at level 3 of testing, the
alertness was found to be related to the duration of eye closure of the sample group each time it was detected. The alertness would stop when the sample group opened their eyes again. And from the results of all 4 levels, the average EAR was 0.17 and 0.15, which is consistent with the threshold set in the program that if there is blinking or close your eyes, the EAR value would be less than 0.20. Therefore, EAR is an important tool for assessing the status of the sample group. If the sample group has normal conditions but the number
of eye blinks is below the specified threshold, the program will count it as drowsiness.
