Research And Application Of Fatigue Monitoring And Behavior Recognition Algorithm Based On TDA2x

Fatigue driving,as the primary factor causing traffic accidents,seriously affects the safety of drivers and passengers.Therefore,the research and development of driver monitoring system has become an important topic in the automotive industry.The fatigue driving monitoring system can monitor the driver’s driving status in real time and give an alarm when the driver feels tired,so as to effectively reduce the occurrence of traffic accidents.However,the current market of a variety of fatigue monitoring systems,it is difficult to balance the hardware platform in the relationship between computing power and cost,power consumption,resulting in fatigue driving monitoring products based on deep learning is difficult to land.In order to solve this problem,a behavior recognition and fatigue monitoring and early warning system based on deep learning is implemented by using low-cost vehicle regulation level chips and sensors.In this paper,the low-cost TDA2 HG chip is used as the hardware platform,the vehicle regulation camera and near-infrared sensor are used as the image sensors in the daytime and at night respectively,and an all-weather fatigue monitoring and early warning system is realized.Based on a sparse and lightweight framework,this paper implements a fatigue detection algorithm based on deep learning SSD(single shot multibox detector)network.Using self collected data calibration and training,a model with detection accuracy of more than 94% is obtained.The model can infer 3.74 frames of 768 * 320 images per second on the evaluation board EVE.The fatigue detection algorithm aims at the EAR(eye aspect ratio)of the driver’s eyes to judge the driver’s fatigue state,and its effectiveness is proved by experiments.At the same time,aiming at the problem of low recognition rate at night,this paper uses near-infrared image as training sample,and combines median filter and wavelet transform to remove noise.Finally,according to the functional requirements of the system,the development process,data flow construction process and algorithm implementation process of the behavior recognition and fatigue driving monitoring system are given on TDA2 HG SoC platform.Through the actual test analysis and further optimization,the system resource occupancy rate and SBL boot time are reduced,the real-time performance and stability of the system are effectively improved,the design requirements are met,and the goal of developing fatigue driving monitoring system on low-cost SoC system is realized.