| BackgroundNowadays,fatigue driving is the primary cause of traffic accidents,bringing huge losses to the society and the people.Many traffic accidents can be avoided if the drivers' fatigue can be detected beforehand.Therefore,this topic is of great social value.Face recognition is the most convenient fatigue detection method currently which can be done with no interruption to drivers since obvious physical changes,especially with the eyes,can be exhibited when in fatigue.In this technology,eye features are the most important features that should be extracted,which have a direct influence on the result of face recognition.PurposeMake software based on face recognition technology,monitor human fatigue through the PERCLOS standard and to give an alarm when the fatigue threshold was reached.MethodsIn this research,gray projection and Hough transformation are firstly applied to find merely one single method of face recognition does not suffice to establish a fatigue driving detection system.Then based on the comparison of three integrated recognition plans provided by some technology companies,a Baidu-based plan is chosen to obtain the coordinates of key points in human face.Next,through Multi-sample single picture,One-sample continuous picture,Multi-sample continuous picture to calculate the PERCLOS value based on model simplification algorithm and general reference determination.Based on that,more effective PERCLOS threshold values are attempted.Results1.The program is completed and get the data of key points in human face.2.99% multi-sample single image can be recognized.3.Data of single-sample continuous pictures could make a model of algorithm and find general reference determination.4.The multi-sample continuous picture data meets the expected PERCLOS standard results.ConclusionThe fatigue monitoring system is feasible based on face recognition technology.Under the P80 standard,when P=0.4 is more consistent with fatigue Driving monitoring. |
PDF Full Text Request