Study On Fatigue Driving Detection Method Based On Combined Features

In modern society,the popularity of the automobile promotes the development of the transportation industry and makes people's lives more convenient,but the frequent traffic accidents seriously threaten the safety of drivers and passengers.What's worse,drivers tend to ignore the physical fatigue feelings and continue to drive,which greatly increased the risk of dangerous occurrence.So,a real-time monitoring of the state for fatigue driving and sending a timely warning are vital to the driver's safety and to avoid traffic accidents of the vehicle.Relying single feature to detect the fatigue driving is easily affected by the environment,which leads to a failure effect and a poor reliability problem of the fatigue detection.To overcome the problem,the fatigue characteristics of eyes and yawn-mouth are combined to detect the fatigue driving in this paper.This dissertation mainly studies the following contents:(1)Research and simulation analysis of the face detection algorithm.The principle of several commonly used face detection algorithms is described and focuses on the theory and process of using Adaboost algorithm to locate the face,then analysis the principle and process of face location relying on the principle of skin color segmentation,and the simulation results of the two methods show that the former method has stronger adaptability and more accuracy.(2)Research and simulation analysis of the eyes location algorithm.Aiming at the problems of eyes positioning are easily influenced by the light and the closed eyes,a method of combine Adaboost algorithm and PCNN segmentation is studied to achieve the precisely eyes positioning.The simulation results show that this method can achieve a good position effect of the human's eyes weather the eyes are closed or not,what's more,it can avoid the spectacle-frame interference to extract eyes' features.(3)A improved algorithm of the mouth location and contour feature extraction.To overcome the lip-like color information interference to the mouth location and the lip thickness interference to distinguish the state of opened and closed mouth.Combine the YCbCr skin-color model and the LAB color space,then use the Gabor transform to extract the mouth contour features.The simulation results show that this method can resist lip-like information as well as the thickness of the lip interference and realize a accurate mouth positioning and yawning discrimination.(4)Fatigue driving detection based on the fusion feature.Analyze the characters of eyes then test the fatigue driving detection according to the PERCLOS method;Analyze the characters of mouth then use the frequency characteristics of yawning mouth to detect fatigue driving,then test the fatigue driving using the fusion feature.By comparing the three methods,the results show that the fusion of multiple features can play a complementary role between different features,and can increase the applicability and accuracy of the fatigue detection system.