Vehicle Safety Monitoring System

In recent years,the smart car industry has developed rapidly.Traditional car manufacturers,new-generation smart car manufacturers,and a large number of Internet companies have entered the smart car industry with their own advantages.Among them,the research and development of automatic driving and smart cockpit are the most popular,but driving safety should be given enough attention as the prerequisite of everything.From the analysis of the occurrence factors of traffic accidents,this paper extracts two major demand backgrounds,one is the monitoring,recording and warning of the vehicle’s own operating conditions,and the other is the judgment of the driver’s improper driving behavior,such as fatigue and distracted driving.Based on the above,the topic of this paper starts from the vehicle’s own driving conditions and the driver’s safe driving behavior,and designs a multi-terminal collaborative safe driving monitoring system.The main work of this paper is focused on the embedded vehicle terminal based on ARM-Linux,and the cloud system is used as the management center,and the mobile phone terminal is used as the function expansion.First of all,this paper analyzes and organizes the related products of automobile black box and the research status of distraction detection,and summarizes the key technologies involved in the vehicle safety monitoring system in this paper.Secondly,the hardware circuit of the vehicle terminal is designed and realized based on the requirements,which lays the foundation for the realization of subsequent functions.After that,realize the driving recording subsystem function on the vehicle terminal.Realize real-time collection and standardized storage of driving data,use MQTT and XXTEA encryption to realize cloud parameter reporting function,design reliable UDP video monitoring function,and realize OTA upgrade function.In addition,on the driver’s safe driving behavior monitoring subsystem,the fatigue monitoring based on the key points of the face to calculate the aspect ratio of the eyes and mouth is realized.Innovatively proposed and experimented with a face gaze estimation model based on channel hierarchical convolution and Mobile Net lightweight processing.The estimation errors on the Gaze360 and MPIIGaze gaze estimation datasets are 12.23° and 4.38° respectively,and the network calculation GFLOPs is only 0.388,implementing a lightweight line-of-sight estimation model suitable for embedded terminals.In addition,this paper migrates the line of sight estimation to the driving environment to judge the driver’s attention area,and achieves a classification accuracy of 96.78% on the LISA Data dataset.Finally,it is deployed on the vehicle terminal and its effect is actually verified.Finally,this paper integrates and completes the function and effect test of the whole system,and verifies the validity and reliability of the function.The vehicle safety monitoring system designed in this paper not only has strong practical value and grounding,but also provides a reference for subsequent smart car safety monitoring terminals and intelligent management.