Research On The System Of Fatigue Driving Monitoring Based On IR Frame Difference And Multi_features

According to the statistics from the communication departments,consequent occurrences of traffic accidents brought us not only enormous economic cost,but also deep psychic damage,resulting in the largest death of population in the world.There are hundreds of reasons that cause traffic accidents,however,among the reasons driving in fatigue is one of the most important.So,if the drivers are timely warned in the beginning of fatigue,the probability of accident happenings will be greatly reduced.In this case,monitoring and preventing drivers from driving in fatigue effectively is meaningful for decrease traffic accidents and mortality rate.By analyzing the current studies from home and abroad,this thesis designed an embedded driver fatigue real time monitoring system based on Difference picture IR and image processing.The system has small size and can be installed easily,meeting the requirements of all.climate working and real time monitoring.The main work is as follows:(1) Analyze the reasons of driver fatigue and the behavior procedure of driving tiredly from the point of physiology and praxiology respectively.And the methods using physiology,behavior and vision to monitor fatigue are also analyzed. Additionally,the way of using individual features extraction to fatigue monitoring based on video and image processing is emphasizedly studied.(2) Propose an algorithm that can locate pupilla on the basis of Difference picture IR and image processing.According to the algorithm,the difference is first carried out on the odd and even frame of image,then smoothing and wave filtering are conducted on the images,after that binarization is used to do image partition by selecting proper threshold,finally pupilla location begins to execute.This algorithm is fast and accurate for eyes location. (3) Propose a method using template matching to extract the stretch of eyes' status.This method,which carries out direct extraction of the eyes' status by template matching,has quicker extraction than other conventional methods that first extracts the eyes' features before extracting the eyes' status.(4) Make use of the template update and Kalman wave filter to track the extracted features.By the template update,the matching accuracy can be improved; and the track speed can also be lifted through the Kalman wave filter.(5) Propose a method that uses the PERCLOS value,nodding frequency and yawn number and duration to monitor the fatigue status.Synthesizing multi_features to define fatigue,the method improves accuracy of monitoring.