The Research And Implementation Of Fatigue Driving Recognition Algorithm Based On Multiple Features

Road traffic accidents caused by fatigue driving have seriously threatened people's lives and property.How to detect the driver's fatigue level in advance and give early warning to avoid traffic accidents fundamentally has become a hot spot in the research field of fatigue driving detection technology.In this paper,after analyzing and summarizing the existing driver fatigue detection methods,a fatigue identification method based on facial multi-feature weighted sum is proposed.The main research work is as follows:1.Image preprocessing and face detection and tracking.Since the captured video images may be contaminated with various types of noise in different degrees during the storage process.In order to ensure the accuracy of face detection,image denoising and illumination compensation are processed in advance.Then AdaBoost face detection algorithm based on Harr-like is used to detect faces.After the face in the image is detected,the facial region is tracked in real time using the tracking algorithm of discriminating scale space.2.In view of the state change of human eyes in fatigue,we propose an open and closed eye state recognition method based on SVM multi-feature fusion.Firstly,a cascade regression tree algorithm is used to locate the feature points of the detected or tracked face,and the human eye region is located according to the position of the human eye feature points.The aspect ratio EAR of human eyes was calculated from12 feature points of human eyes,and the state of eye opening and closing was identified according to the EAR.An adaptive domain method is used to calculate the cumulative difference of black pixels in the binary image of human eyes.Finally,the eigenvalues obtained by the two methods are used as input parameters of the SVM classifier for model training,and the trained model is used for the recognition of the open and closed state of human eyes.3.For the identification of the fatigue state of the mouth,the aspect ratio MAR of the mouth was mainly calculated based on 10 characteristic points of the mouth,and the yawning state was identified according to setting the MAR threshold.In the head fatigue state,the head frequency in the two-dimensional vertical direction is calculated according to the head movement.The position of the two center points in the positioned left and right eye area was calculated,and the midpoint of the twocenter points was taken as the head position motion analysis point.The change of the vertical coordinate y of the detection point with time was used to calculate the nod frequency within the time period.4.The fatigue identification model was established according to the fatigue characteristics of eyes,mouth and head,and a fatigue identification method based on multi-feature weighted sum is proposed.The duration and closed eye frame ratio of eye fatigue parameters were extracted according to the condition of eye opening and closing.Blink frequency can be obtained by blink detection.The number and duration of yawning were measured by yawning test.The head frequency is obtained by analyzing the head motion state.These indicators are weighted and summed to evaluate the fatigue level of drivers and give corresponding early warning.Experimental results show that the method proposed in this paper has high accuracy and good real-time performance.