Research On Fatigue Detection System Based On Multi-Source Information Fusion

Fatigue driving is an important incentive in traffic accidents.Research on driver fatigue detection technology has great theoretical and practical significance.To overcome the susceptibility of the detection scheme based on single feature or similar feature fusion in existing fatigue detection to interference,this thesis based on the theory of multi-source information fusion to improve the existing technical means of fatigue detection,based on the human body A set of fatigue monitoring systems for drivers is designed based on the fact that there are many types of fatigue characteristics under fatigue.For this reason,the thesis mainly studies from the following aspects.The advantage of the detection system using multi-source information fusion technology over single or similar fatigue feature detection schemes is discussed,which guarantees the theoretical basis in system design;a scientific experimental paradigm is designed,combined with simulated driving experiments from machine vision,Physiological signals,behavioral characteristics of the three categories selected effective PERCLOS value,eyelid opening degree,average blink duration,blink times,heart rate,pulse,duration of effective driving behavior and other seven characteristics of the driver's fatigue status of the study;The PVT theoretical design experiments for the determination of human fatigue state in neuroscience are used to quantify the fatigue state of the driver.The fatigue analysis system is modeled by a fusion algorithm such as regression analysis,SVM state machine,fuzzy mathematics,and neural network.Combined with the designed simulation experiment,the performance of the detection system under different models was analyzed.Finally,the development process and results of the detection system were explained and demonstrated.Simulated experimental data analysis shows that the trends of effective PERCLOS value,average blink duration,and duration of no effective driving behavior are positively related to the driver's fatigue status,and the trends of eyelid opening,blink times,heart rate,pulse,etc.It is negatively correlated with the fatigue status of the driver;the fatigue characteristics of the system are selected properly,and the accuracy of the system detection under the neural network model is 85%;compared with a single or similar fatigue feature detection scheme,the combined system has higher detection efficiency,stronger anti-interference ability.The software and hardware debugging results show that the modules of the system work normally and the detection accuracy in the actual scene reaches 80%,which has great research significance and wide application prospects.