Research On Fatigue Driving Detection Technology Based On Head Feature And Attitude Estimation

In recent years,with the rapid development of the automobile industry,people's demand for fast,comfortable and convenient living has been continuously improved,and self-driving travel has become the first choice for modern families.However,the increasing car ownership will not only increase traffic congestion caused by road burden,but also directly lead to an increase in traffic accident rate.Among them,the number of traffic accidents caused by fatigue has been increasing,and the serious consequences brought about by it have gradually surpassed the causes of accidents such as drunk driving and overloading,so fatigue driving has begun to attract people's attention.Therefore,in order to reduce the occurrence of vicious traffic accidents and create a safe and comfortable driving environment,it is important to study effective fatigue driving detection technology.At present,most of the fatigue driving detection methods are based on extracting a single characteristic index,the detection rate is not high and the environmental requirements are strict.Based on this,this paper proposes a new method based on head feature and attitude estimation for multi-feature fusion to quickly detect driving fatigue.The method which is low in cost does not require external equipment,and ensures high precision with good robustness.Considering that the human eye is susceptible to occlusion,this article adds head pose parameters to assess fatigue.In this paper,the driver's face features are extracted based on Histogram of Oriented Gradient feature combined with the Conditional Local Neural Fields algorithm.Then the head pose estimation is implemented based on the EPnP algorithm based.Then four evaluation indicators are defined and calculated: head forward Nodding frequency,abnormal head frequency,blink frequency,PERCLOS value.Finally,the multi-feature index fusion method based on rough set theory is used to realize fatigue driving evaluation.The method of this paper has proved its feasibility through experiments.The research content of this paper mainly includes the following aspects:1.Extracting facial features in real time is a prerequisite for implementing this method.In this paper,the head state of the driver is monitored in real time by the camera.Firstly,the face detection is realized based on the Histogram of Oriented Gradient feature.Based on the acquisition of the face,the face feature points are located and extracted based on the Conditional Local Neural Fields algorithm.The coordinate information of the face feature points obtained in this paper will be the input to calculate the evaluation index of each fatigue state.2.Driver head attitude estimation is a process of conversion from a two-dimensional video image to a three-dimensional space.Based on the representative 12 face feature point 2D coordinates,standard 3D face model coordinates and camera internal parameters obtained by camera calibration,the driver's head posture is estimated based on the EPnP algorithm.The method of this paper can obtain the three-dimensional deflection angle of the driver's head posture in real time.3.The particularity of the driving environment leads to a single feature that is not enough to reflect the degree of fatigue.Therefore,this paper proposes a multi-feature index fusion method: based on facial feature coordinates and three-dimensional head pose,introduce the PERCLOS method,define and calculate four evaluation indicators: PERCLOS value,nodding rate,head abnormal rate and blink rate.The evaluation thresholds of each feature based on fatigue degree are set.Finally,the multi-feature index fusion method based on rough set theory is used to evaluate the fatigue driving degree.