Research On Driver Fatigue Detection Algorithm Based On Active Shape Model And Multi-clues Fusion

Fatigue driving in the car safety is a powerful invisible killer,paralysis,at the same time,to human life and property security has brought great hidden danger.In order to reduce the damage caused by fatigue driving and improve the accuracy and robustness of driver fatigue detection,this paper proposes a multiple-clues fusion fatigue detection algorithm based on active shape model.The algorithm uses the camera to shoot the driver's face image,and then uses the Haar-like features of the cascade Adaboost algorithm for face location.Locating feature points of eyes,mouth and head by active shape model according to the extracted face information,then obtain the characteristic parameters related to driver fatigue.Finally,the driver's fatigue degree is judged by the adaptive neural inference system.This paper mainly focuses on the following aspects:(1)Taking full account of the real time of the algorithm,the Haar feature is used to detect the position of the human face quickly by using the cascade Adaboost algorithm and revise the result by skin color.The initial shape of the active shape model is constrained by the face detection algorithm,and this can improve the computation speed of the active shape model.(2)In view of the difficulty of extracting the parameters of the eyes,mouth and head during the fatigue testing.The active shape model is used to locate the feature points in the range of the human face,and the parameters of the eyes,mouth and head can be extracted accurately by 12 feature points.(3)In order to improve the robustness and accuracy of the algorithm,a multi feature fusion based fatigue detection algorithm is used.On the basis of the parameters of the driver's eyes,mouth and head,four fatigue characteristic parameters,such as PECLOS,AECS,yawning frequency and nodding frequency,were obtained.(4)A classification algorithm based on adaptive neuro fuzzy inference system is designed by using the PERCLOS,AECS,yawning frequency and nodding frequency as the input values.The driver's fatigue degree is divided into three grades:awake,fatigue and severe fatigue.The data collected from the experiment is used to train the classification system,which solves the problem of difficult to quantify the degree of fatigue.