Research On Fatigue Driving Detection Method Based On Random Forest And Multi-source Information Fusion

The WHO statistical report states that approximately 1.35 million people die from road safety accidents worldwide each year.Among these major road safety accidents,traffic accidents caused by fatigue driving account for an extremely high proportion.It is necessary to design a system that can be easily installed in the vehicle to detect the fatigue state of the driver in real time.When the driver becomes fatigued,an alarm is issued to prompt the driver to stop in time to avoid the loss of life and property caused by fatigue driving.Most of the existing fatigue driving detection methods only use a single signal to detect the driver’s fatigue state,which is susceptible to interference.When the actual detection environment is not so ideal,sufficient accurate data cannot be collected,which will cause large errors in the detection system and false alarms.This paper designs a multi-source information fusion detection method.By building a signal acquisition platform,the circuit design uses multiple high-precision miniature sensors to synchronously collect the driver’s breathing,heartbeat,pulse,grip strength and other signals.When the collection of a certain signal is deviated,the acquisition of another signal is not disturbed,and then the information is processed by filtering,Fourier transform,etc.,to establish a multi-source fusion fatigue driving state data set as the detection basis,avoiding the defect of poor anti-interference performance of a single signal,which greatly improves the accuracy of fatigue driving detection.Compared with classification algorithms such as SVM(support vector machines)and GBDT(Gradient Boosting Decision Tree),random forest can balance the error of the data set,is insensitive to missing values,can perform parallel operations,and has a high classification speed,which meets the requirements of certain errors in signal collection and high real-time requirements in fatigue driving detection.Through the design model RF(Random Forest),the established multi-source fusion data set is learned and trained,and the parameters are continuously adjusted to obtain the best fatigue driving state detection model.Experiments are carried out on a single data set and a multi-source information fusion data set.And the performance of GBDT and SVM models on multi-source information fusion data sets is compared.The experimental results verify that the algorithm model based on random forest and multi-source information fusion can accurately detect the fatigue driving state.The best detection accuracy of the random forest algorithm model after multi-source information fusion reaches 89.16%,which is compared with a single data set.It is increased by about 15% compared with GBDT and SVM,its training speed is faster and the accuracy is higher,which meets the expected real-time and highprecision requirements for fatigue driving detection.