Research On Recognition Method Of Automobile Driving Fatigue Based On Machine Vision

According to statistics,90% of traffic accidents are caused by human factors,and the traffic accident rate caused by fatigue driving is particularly significant.Therefore,the Ministry of transportation requires terminal equipment to detect fatigue,which can monitor the continuous driving time of drivers,identify blinking,yawning and fatigue state.However,in the existing studies,the selection of drivers’ eye features is relatively simple,and the recognition of drivers’ nodding is not considered,which leads to frequent false detection and missed detection.In addition,it only realizes the driver fatigue recognition function,and does not retain the driver fatigue data.Therefore,this paper realizes the face recognition algorithm,and based on this,completes the function of continuous driving duration monitoring and saving driving data.Then,through five kinds of eye feature parameters,mouth feature and nod feature,a comprehensive fatigue recognition model meeting the requirements of relevant laws and regulations is realized.Firstly,this paper detects the driver’s face based on mtcnn algorithm,extracts the128 dimensional feature vector of the current driver through residual network,calculates the Euclidean distance between the face registries,and recognizes the driver’s identity information.Based on this,the driver’s continuous driving duration detection function is completed.Then,the fatigue driving experiment is designed on the driver simulation platform,and the relevant experimental data and subjective fatigue evaluation are collected.On this basis,through the statistical method,the validity of five eye characteristic parameters,such as PERCLOS parameters,eye aspect ratio,is verified,and the corresponding calculation method is proposed.Then the recognition algorithm of yawning and nodding actions is realized.In addition,the feature compensation method based on Euler angle and distance mapping solves the feature error caused by shooting angle and shooting distance.Finally,the eye fatigue recognition model is constructed based on support vector machine algorithm,and a comprehensive fatigue recognition model is constructed by combining yawn recognition model and nod recognition model.Then,the verification of each model is completed on the experimental data set.Experiments show that the model can meet the requirements of relevant laws and regulations.Finally,the rk3399 high performance motherboard based on firefly is used as the hardware development environment to integrate the fatigue identification algorithm.