Research And Development Of Driver Monitoring System Based On Fatigue Detection And Distraction Detection

With the gradual increase in car ownership,the number of global traffic accidents has remained high,30% of which are caused by drivers’ fatigue driving and distracted driving,especially in major traffic accidents.Today,when autonomous driving technology cannot be widely popularized,it is an inevitable trend to give priority to the development of assisted driving.Therefore,research on the Driver Monitoring System(DMS)can effectively avoid or reduce the occurrence of traffic accidents,which has great practical significance to protect the life safety of drivers and occupants.Based on the above knowledge,this paper studies the driver monitoring system based on fatigue dectection and distraction detection to monitor the abnormal state of the driver to ensure driving safety.The main content of this paper is as follows:Firstly,based on the collected driver’s facial image,preprocessing such as grayscale,Gaussian filtering,and contrast enhancement is completed according to the needs of the face detection algorithm.The various face detection algorithms are summarized,and the training of two face detection classifiers based on Haar+Adaboost and HOG+SVM is completed.The two methods are simulated and tested through two datasets,and then selecting the method based on HOG+SVM for face detection.Secondly,the ERT algorithm is used to complete the positioning of the facial feature points.Based on this,completing the extraction of driver’s eyes(PERCLOS value,blink frequency,etc.),mouth(yawn frequency)and head(head posture)fatigue features.Finally,a fatigue detection method based on multiple fusion features is proposed to realize real-time driver’s fatigue detection under various complex conditions,with a detection accuracy rate of 92.0%.Then,the traditional pupil center positioning algorithm based on image gradient is improved to realize the accurate positioning of the pupil center.According to the driver’s pupil center coordinates and head posture,the driver’s gazing area is estimated.Through the image of the road ahead,completing the detection,screening,and area identification of dangerous targets.Based on the driver’s attention area and dangerous targets’ area,a hierarchical detection method of distraction is proposed to detect the driver’s distracted state in real time,with a detection accuracy rate of 91.6%.Finally,the two sub-modules of fatigue detection and distraction detection are combined to form a complete driver monitoring system.The entire system is tested through actual vehicle tests.The test results show that the driver monitoring system proposed in this paper has excellent performance and can accurately judge the fatigue and distraction status of the driver.In the case of failure of general detection methods,such as large head deflection,poor lighting conditions,wearing glasses and sunglasses,the driver monitoring system proposed in this paper can still work efficiently.Because the two core modules have a certain degree of redundancy for the driver state detection process,when a single feature is disturbed and cannot be accurately identified,it can also be accurately identified based on other features,which improves the robustness of the system.