Research On Fatigue Driving State Detection Based On SDM

With the increase in the number of car ownership,traffic accidents have also increased.Among them,traffic accidents caused by fatigue driving are relatively large and can't be ignored.This paper makes a deep research on vision based fatigue driving status detection method.A multi-index linear fusion fatigue driving state detection method based on Supervised Descent Method is proposed to achieve accurate and rapid fatigue driving state judgment.The main work of this paper includes the following aspects:?.A fast face detection method is proposed.Firstly,the skin color segmentation algorithm is used to locate the initial position of the face region quickly,and then the AdaBoost algorithm is adopted to complete the secondary positioning.The experimental results show that the proposed method improves the real-time performance of the AdaBoost algorithm,and also compensates for the insufficiency of the skin color segmentation method.?.The SDM algorithm is introduced into the positioning of face feature points to get the more accurate positioning of face feature points.When the face is occluded and the head posture changes obviously,the feature point positioning will drift.In response to this deficiency,the SDM algorithm is used to obtain the face feature points position more accurately.The algorithm was tested on the CAS-PEAL-R1 dataset.The experimental results show that the SDM algorithm can locate the face feature points accurately in the case of different backgrounds,distances,illumination and complex expressions,and can provide more accurate data for the fatigue parameter extraction.?.The POSIT algorithm is applied to realize accurate estimation of the head posture.Based on the location of the two-dimensional face feature points,the three-dimensional head model and the camera internal parameters,pitch,yaw,and roll three angles that characterize the head pose can be calculated iteratively.Tests were performed on Pointing'04 and CAS-PEAL-R1 two datasets.The experimental results show that the accuracy of the key angles(pitch angles)used to characterize fatigue in the algorithm is 94.29% and 95.33% respectively,which can satisfy the fatigue state detection.?.A multi-index linear fusion method is adopted to judge the fatigue driving state,it can characterize the fatigue state more comprehensively and accurately.Three fatigue index such as continuous closed eye frames,continuous yawning frames,and head offset duration frames are used to perform multi-state linear weighted fusion,in order to calculate the fatigue index and characterize the fatigue state.The experimental results show that the accuracy of the multi-index fatigue detection method in this paper is higher than that of a single index fatigue detection method.