Research On The Detection Method Of Driving Fatigue Based On Deep Learning

As an effective way to improve social safety and well-being,driving fatigue state detection improves the reliability of the driver’s passive safety.In practical applications,fatigue can be found in time and early warning fatigue,thereby effectively preventing accidents and ensuring The safety during the journey has strong practical significance.In this paper,the fatigue detection research analyzes the driver’s state while driving the vehicle,and mainly focuses on the driver’s facial details,and has successively carried out research on face detection and deep neural network to distinguish fatigue.In the traditional face detection method based on Adaboost,in continuous training,the proportion of positive cases passing is getting higher and higher,while the proportion of negative cases passing is getting lower and lower.Finally,a large number of negative cases are filtered out,with higher Detection accuracy,but time consumption is relatively large.This paper adds three new Haar-like feature blocks and improves the face detection implemented by the Adaboost algorithm of the weak classifier connection method.Finally,by comparing with the traditional Adaboost face detection experiment,according to the detection accuracy,speed and robustness comparison in different external environments,the superior effect of the improved algorithm is reflected.The traditional driving fatigue detection method is based on the physiological signals of the driver,such as monitoring the physiological information of the cerebral cortex,the second is based on vehicle manipulation information,such as calculating the steering wheel rotation angle and other vehicle information,and the third is based on the driver behavior information.Such as the facial features of the driver.After analyzing the current research status at home and abroad,it is found that the length of time of opening and closing of human eyes and the frequency of blinking(BF)are the main signals of fatigue.Therefore,this paper selects the driver’s eye characteristics as the fundamental criterion for fatigue judgment.In the process of fatigue discrimination,the Re-Dense Net model of improved Dense Net is adopted,the loss function of the network is improved,and the connection mode and number in the network are changed to achieve the effect of increasing the detection speed and reducing the amount of calculation,so that the network has a stronger classification and recognition ability.Accurately divide the state of driver fatigue and non-fatigue.At the end of the experiment,use the method in this article to experiment on CAS-PEAL and self-built data sets.The experimental comparison results show that the method in this paper has a good detection effect on different data sets,and proves the validity and correctness of the proposed experimental method.This undoubtedly provides a strong impetus and forward force for driving fatigue warning research.Artificial intelligence is developing rapidly,and deep learning research is hot.It is imperative to use deep learning to carry out the research of driving fatigue state detection.The neural network will improve the detection performance with the continuous increase of the level,but the excessive increase of the level will bring about problems such as gradient dispersion.In order to avoid this problem,this article chooses Dense Net is the basic network for driving fatigue detection.During the experiment,the network settings are continuously improved,and the network finally obtains good detection performance.