Research On Adaptive Mechanism For Driver Fatigue State Detection Based On Deep Network

Driver fatigue is a major cause of traffic accidents.Relevant researchers have conducted extensive research on fatigue detection problems and have achieved a series of research results.However,it is difficult to detect the fatigue state accurately and reliably due to the difference of individual characteristics.This research focuses on the problem of insufficient adaptability of the fatigue detection model,based on the analysis for the fatigue state change rules and detection mechanism,an adaptive driver fatigue detection method based on deep network is proposed which can effectively solve the key problems in the process of related on-board systems development.The main research contents include the followings:(1)Adaptive mechanism for driver fatigueThe analysis and representation of individual characteristics is the key to accurately detect fatigue state and effectively improve the adaptability of the detection model.Therefore,based on the analysis of individual characteristics,this research proposed an adaptive strategy for driver fatigue state detection model by analyzing the coupling between driver characteristics and fatigue detection model performance,which lays the theoretical foundation for the construction of adaptive model of fatigue detection.(2)Fatigue detection model construction based on deep networkThe generation of driver fatigue has typical time-series characteristics,the reliable detection of fatigue state is related to a variety of drivers’ apparent characteristics,a reasonable identification model is the key to fatigue detection.Therefore,based on the analysis of the dynamic generation characteristics of the driver fatigue state,this research constructs a cascaded CNN-LSTM fatigue state detection model through the analysis of the fatigue state and the apparent characteristics of related drivers,furthermore,the optical flow feature is introduced to improve the reliability of model,which lays the foundation for the fatigue adaptive detection model construction.(3)Fatigue adaptive detection model constructionAccording to the research purpose of this research,it is the key of model adaptive detection to introduce the individual characteristics into the fatigue detection model.Therefore,based on the analysis of individual characteristics and adaptive mechanism,this research proposed an adaptive detection model of fatigue state based on deep network by improving the framework of fatigue detection model,and then made a verification on the model with relevant experimental data.This research carried out related research work on the analysis of the driver fatigue state adaptive detection mechanism and the adaptive detection model construction.Among them,the focus of the research is: aiming at the core problem of insufficient adaptability of the fatigue detection,a self-adaptive mechanism and construction method are deeply explored to solve the key mechanistic problems of related on-board system development through analyzing individual characteristics and the driver fatigue detection mechanism.