| Driver fatigue is an important cause of road traffic accidents.Fatigue makes drivers less alert to changes in the external environment,reduces their ability to maintain safe driving,increases reaction time to complex stimuli and increases the likelihood of driving errors.Effective prevention and control of road traffic safety from the driver’s perspective has attracted increasing social attention.In order to solve the problem of fatigue driving,it is important to explore the mechanism and behavioural causes of driver fatigue driving,as well as to implement driver fatigue state detection,both of which are important.In this thesis,based on the selection of certain test samples,these two parts of the research are carried out: the analysis of the factors influencing the behavioural intention of car drivers to drive fatigued based on the theory of planned behaviour;and an improved LBP algorithm for fatigue feature extraction and fatigue state detection based on the driver’s driving fatigue facial features.The research content of this paper is summarised as follows.1.By describing the serious consequences of fatigue driving behaviour in traffic safety,the current research models are compared through an overview of the current sttus,research results and development trends,and the traffic behaviour of drivers is studied from the perspective of social psychology,and a mathematical model of the structural equations of drivers’ fatigue driving behaviour is established.Based on the three basic variables of the Theory of Planned Behaviour(TPB),the explanatory variable of risk perception was added to analyse the factors influencing the intention of fatigue driving behaviour.Using the idea of variable analysis of structural equations,a structural equation of drivers’ fatigue driving behaviour was constructed based on the Extended Theory of Planned Behaviour(Ex-TPB)and studied from the psychological perspective of drivers.The empirical analysis found that the driver’s risk perception has the greatest and positive influence on the behavioural intention to drive fatigued,while perceptual behavioural control has the second highest influence.2.By outlining various image feature recognition technology points,we analyze the applicability and development of LBP texture extraction algorithm in the field of facial feature recognition since home and abroad.An improved algorithm based on LBP features is proposed and combined with the texture feature extraction algorithm based on spatial relationship,which introduces spatial regularity features in LBP coding,thus effectively enhancing the feature expression of LBP.To avoid the problem of local difference information loss and strengthen the ability of describing partition of image content,the face features are specially extracted in chunks,which constitute chunked circular LBP features.To verify the effectiveness of the improvement,the improved points are compared with the basic LBP feature operator,the classical HOG feature detection and the MTCNN detection method pair in deep learning for control experiments.The experiments show that the improvement points proposed in this paper can effectively improve the recognition rate of LBP features.3.To solve the problem of fatigue detection of driver’s facial features,the improved LBP algorithm is proposed to extract the driver’s facial feature texture as the input of the convolutional neural network,and the model is trained by the improved LeNet-5 network structure.In order to achieve the best learning performance of the convolutional neural network,anti-overfitting strategies such as Relu activation function,Dropout random rounding and BN normalization are used.The model is trained by using the convolutional neural network structure.The results of the simulation experiments on the self-built driver fatigue image test set show that the accuracy of the model is 93.52%,and the fatigue detection model has good recognition accuracy and generalization ability. |
