Research And Implementation Of Driver Fatigue Detection Method Based On Multi-feature Fusion

Fatigue driving detection is a singificant research field in various scientific research institutions.Among the many research results,the detection based on driver’s facial features is the significant research in the field of current fatigue driving detection.Nonetheless,this method has some problems,such as the great influence of light on detection accuracy,the difficult balance between real-time performance and accuracy,and the greater errors of single fatigue driving judgment factor.Therefore,this paper proposes a driver fatigue detection algorithm that can effectively improve the detection accuracy under weak light conditions,multi-feature fusion judgment,and good balance between real-time performance and accuracy.The research contents are as follows:(1)The fatigue driving of the driver mainly occurs at night.when the light is relatively dark,the features of the driver’s eyes and mouth are unconspicuous,which influences the recognition accuracy.So before the image enters the face detection and extraction,it should be processed to increase the quality,So as to improve the subsequent eyes open state and mouth open state recognition accuracy.(2)Retina Face algorithm is used to accomplish the facial detection of drivers and abstract facial area images.Then I introduce the face key point localization algorithm,uesd to accurately obtain the face contour,eyes,mouth,nose and other 68 key areas of the face coordinate information,providing the driver’s head posture estimation for coordinate of key points.(3)The basic process of convolutional neural network model construction is analyzed.Then,a convolutional neural network model is designed to distinguish the state of driver’s eyes and mouth based on the SSD network model and combined with the application scenarios in this paper.Then the model is trained with a self-made data set to obtain the most appropriate key parameters such as learning rate and batch processing size.(4)Fusion of parameters commonly used in fatigue driving detection,such as PERCLOS parameters,yawn parameters,and abnormal head posture parameters,are combined with the actual situation in this article to give the parameters suitable for the fatigue driving detection method,such as eyes during fatigue driving.Integrating these fatigue driving judgment basis information to establish a fatigue driving recognition model can effectively improve the robustness of the system.After proper dataset testing,the fatigue driving detection system in this paper has certain practical application value.