Research On The Application Of Steering Wheel Angle Characteristics In Driver Fatigue Detection

With the development of road traffic network's construction and the rapid increase of vehicle population,road traffic accident becomes an important problem which seriously affects the development of society.Through collecting and sorting out the statistics of road traffic accident for the past 10 years, we find that the number of road traffic accidents and casualties has declined in the recent years; however there has been an upward trend for the average number of casualties per accident.Among all the causes of road traffic accidents,diver fatigue is the vital one.Therefore,there is an important practical significance for research on driver fatigue.On the basis of comprehensively describing the research status of driver fatigue detection,it shows that most of the researches were only carried out in laboratory or paid less attention to drivers' physiological and psychological feelings.Aimed at the above mentioned problems,a steering wheel angle acquisition device was designed and road driving test was carried out.Through real driving test, angle data was obtained from normal and drowsiness driving state,based on that,it conducted comparative analysis and proposed the standard of classification.Thus data support for driver fatigue detection was provided.Researched on the driver fatigue detection's modeling process which relayed on BP neural network, on the basis of basic theory of artificial neural network,structure of BP neural network,error back propagation algorithm and its improvement.Discussed following issues such as best number of hidden node,pretreatment of data and determination of training algorithm.The model was applied in test as well.The comparative analysis between the experimental data and the simulation outputs shows that the model could detect driver fatigue at the rate of 71.4%while it proved the feasibility and reliability of the BP neural network when used in the driver fatigue detection.Aimed at the shortcomings of the BP neural network,such as easily getting local minima,a combined BP network using Genetic Algorithm was proposed.By using the global searching ability of Genetic Algorithm,the initial weights and threshold of neural network were optimized,thus the network performance was improved.The simulation result indicates that the detection rate has achieved 78.2%, increased by 9.5%.With Genetic Algorithm,the generalization ability of BP neural network was improved and detection precision of the model was further enhanced.