Research And Design Of Driver Fatigue Detection System Based On Embedded System

With the development of society and the progress of modern science and technology,automobiles have played an increasingly important role in our daily life and in all walks of life.In recent years,the number of automobiles in our country has reached a record high,subsequently,the number of traffic accidents is also increasing,and fatigue driving is one of the three major causes of traffic accidents.According to statistics,the proportion of traffic accidents caused by fatigue driving is 14% to 20%,and the proportion of fatigue driving in major traffic accidents is as high as 43%.In order to reduce the traffic accidents caused by fatigue driving and ensure the safety of vehicles and drivers,researchers at home and abroad have never stopped the study of fatigue driving detection,and now has achieved remarkable results.Among the numerous detection methods,the method based on drivers facial features,which has the advantages of accurate identification,low cost and no influence on drivers driving,has been paid more and more attention by scholars.After analyzing the research results of fatigue driving detection at home and abroad,this paper innovatively proposes an end-to-end deep neural network model,which uses a large number of face images for model training to detect the state of drivers eyes and mouth,a real-time and stable driver fatigue detection system based on embedded platform is designed by using PERCLOS algorithm to judge driver fatigue degree.The main work of this paper is as follows:1)Image preprocessing.In this paper,for face feature detection,the convolutional neural network is used to extract the feature data of drivers face image,and the weight parameters are optimized repeatedly.In the process of face image acquisition,the difference of light or the performance of machine equipment will lead to the lack of noise or contrast of face image.In this paper,image grayscale and two-dimensional Gabor filter are used to denoise and preprocess the image.2)Face detection.The deep learning algorithm is actually the superposition of multi-layer convolutional neural network,which convolves the image data,outputs the feature map,then performs the next convolve,obtains the deeper information of the image,and finally describes the target feature with the feature map,the Algorithm chosen in this paper is the deep learning target detection Algorithm SSD with high accuracy and short time,and the SSD is improved based on the CRe LU function and the characteristic pyramid FPN,thus the detection precision is improved,at the same time,the efficiency of network information utilization has also been improved3)Fatigue detection.After training,the convolutional neural network model can judge the state of drivers eyes and mouth easily.The state information of eyes is combined with PERCLOS Algorithm,the state information of mouth is used to judge Yawn,and the fatigue state of driver is detected by combining the fatigue parameters of eyes and mouth.4)Embedded porting.In this paper,we choose information as Itop4412 Development Board,use Qt technology to develop human-computer interface,carry Linux operating system,image processing function through Open CV library.The system is tested in real driving environment.The algorithm is short,the power consumption is very low,and it can be used in fatigue detection basically.