Identification Of Vehicle Fatigue Driving State Based On Machine Vision

Vehicle ownership in our country is growing,and uncoordinated development of road space and traffic demand leads to the deterioration of traffic safety and the traffic environment,which seriously affects people’s lives and property safety.Research shows that fatigue in driving has become one of the main causes of traffic accidents.Therefore,it is of great practical and theoretical significance to identify driver’s fatigue in real-time and accurately and provide timely active safety warning.Existing studies on fatigue driving state recognition mainly rely on eye representation parameters,but only single element features are considered,which cannot accurately identify complex fatigue states.Therefore,a non-invasive and low-cost fatigue driving state identification method based on genetic optimization algorithm was proposed in this paper.The specific work of this study is mainly as follows:（1）Simulated driving experiments and real vehicle driving experiments were designed,and experimental data of 33 subjects during normal driving and fatigue driving were collected.After the experiment,the driver’s fatigue state was determined according to the two-category KSS scale,and the fatigue driving sample database was established.（2）The MTCNN network was improved and applied to face detection,by which localization of face areas in simulated and real driving environments was realized.The Dlib toolkit was introduced to extract the coordinate values of facial feature points,and feature parameters of the driver’s eyes and mouth were collected.The head pose was estimated according to the mapping relationship between the 2D face key points’ coordinates and the 3D face model coordinate.Combined with qualitative and quantitative analysis methods,the data fluctuation of each feature parameter of fatigue driving was analyzed.The fatigue identification model was constructed by using multiple characteristic parameters,by which the problem of poor model stability and generalization ability caused by a single parameter was made up.（3）The data was preprocessed by using a method of Fast Fourier Filtering（FFT）.Method of factor analysis was used to extract main factors as main features of the fatigue identification model.The genetic algorithm（GA）was used to find the optimal smooth factor of generalized regression neural network（GRNN）,and the fatigue driving identification model of GA-GRNN was established.To verify the superiority of the GA-GRNN model,it was compared with KNN,RF,and GRNN fatigue driving identification algorithms.The results of experiments showed that the accuracy of the GA-GRNN model for fatigue driving state identification is as high as 93.3%,the recall is 91.4%,the precision is 92.9%,and the F1 score is 92.1%.In this study,the feasibility of machine vision technology for fatigue driving detection was demonstrated,which can provide theoretical and technical support for the application of early fatigue driving identification for drivers.