پارسا فخري

Rail transportation safety is the most important issue that is always being sought to improve. According to the statistics obtained from rail accidents caused by fatigue and drowsiness, in Japanese railways it accounts for about 17%, in the Federal Railroad Administration of the United States about 41%, and in Europe about 21%. These accidents, given the long working hours of train drivers, need to be managed and prevented. Therefore, a reliable technology for detecting drowsiness of train drivers is needed to prevent these accidents, which can be used to activate audio and visual alarms in case of driver's lack of al‎e‎rtness and prevent accidents. Deep learning techniques have been widely used in recent years to create these strategies. This work focuses on methods for detecting driver drowsiness that use YOLOv7 and YOLOv8 models. Given the better performance of YOLOv8 (97.1%) , the weights trained on it were used to build the program, and the real-time detection features using YOLOv8 models are also shown in this study. In addition, to prevent the program from being fooled, a feature has been added to it using dlib to identify the driver's face. The output of the program has also been eva‎luated and its results with different conditions have been presented.

راه آهن , تشخيص خواب آلودگي , الگوريتم YOLO , يادگيري عميق , پردازش تصوير

Railway , YOLO Algorithm , Deep Learning , Image processing , Fatigue Detection

Parsa Fakhri

Ahamd Mirabadi