Monitoring For Fatigue Driving State Of The Operator

Research and development of high-performance fatigue real-time monitoring technology can effectively determine whether the operator into the state of fatigue.This can not only improve the working efficiency,but also solve the potential safety hazard.The thesis focuses on a non-contact,real-time fatigue detection method based on video,and proposes an algorithm with multiple fatigue characteristics.Firstly,it collects the video through the camera and carries out simple preprocessing.Then,the face area is quickly located by AdaBoost and the face shape model is constructed by ASM,which is used for locating the eye and mouth precisely,and extracting the relevant parameters.And then the operator's head movement is estimated.Based on the above indicators,it establishes the mapping relation between the characteristic space and fatigue space to judge the status with the SVM.The details are as follows:1)According to the AdaBoost and ASM,the precise location of the facial feature points is studied.AdaBoost narrows the range of face,and ASM has a good tracking effect on key points.The combination of them can achieve the positioning of point.2)It analyzes the changes of multiple behavioral states in detailedly,and determines the characteristic parameters of eye,which includes the PERCLOS value,the average closed eye length and the blink frequency.Meanwhile,with the reference of the PERCLOS principle,the definition of PMRCLUS is proposed as the yaw threshold,which is determined by a large amount of experimental data.3)It gives the comparation of the image pose based on POSIT and the space motion based on inertial sensor,and then,it fuses the data with the two methods,which can extract the head parameters with high accuracy.4)It integrates various of parameters,which improves the reliability of system.With building a fatigue monitoring model with multiple types of characteristic parameters,the system can finally determine whether the operator is in fatigue.Experimental data shows that the precision of the system reaches to 86%,which indicates the high accuracy of the test.The research results of the thesis can provide support for the fatigue monitoring technology to the operator in a certain extent.