Research On Body Shape Recognition Of Drivers And Occupants In The Front Row Of Vehicles

In automobile collision accidents,the vehicle occupant restraint system is a key factor to prevent and reduce occupant injuries.However,the traditional occupant restraint system is designed according to the 50 th percentile adult male occupant body shape,and cannot provide the same level of safety protection for other body occupants.Even statistics show that the improper matching of the occupant restraint system will increase the risk of occupant injury.In response to this problem,it is necessary to carry out research on the body shape recognition of the occupants in the car to lay the foundation for the protection of occupants of various body types.However,the current occupant body shape research can only divide the occupants into a limited number of categories,and cannot accurately identify the height and weight of the occupants and provide detailed body shape information.Therefore,this article combines machine vision,volunteer experiments,statistical analysis and other technologies to realize the precise identification of the body shape of the front row of drivers and passengers in the car.The research content of this article is as follows:1 Establish a data set containing Asian faces and preprocess the data set,conduct in-depth research on the principle of deep learning of convolutional neural network,use convolutional neural network to train the processed data set,and obtain age and gender recognize the model and test the trained model to verify its recognition accuracy.2 Carry out volunteer experiments to collect data on the main characteristics of the human body such as height,weight,head height,width between the two tragus,shoulder width,and upper arm length.Use the data obtained from the volunteer experiment to perform regression analysis to establish the regression model of height,weight and the characteristic size of the human body.3 Use the multi-cascade regression tree to perform the key point regression method to establish the entire head feature point extraction model.Based on Dlib and openpose technology,the feature points of the head,shoulders and elbows of the front-row drivers and occupants in the car are extracted,and the size data of the head,shoulders and elbows of the front-row drivers and occupants in the car are calculated with binocular vision.Substituting the above-identified feature size of the human body into the established height and weight regression model to obtain the height and weight information of the front-row driver and occupant in the car.4 Use volunteer data to verify the accuracy of the body shape recognition model for the front row of drivers and passengers in the car,and analyze the body shape parameter recognition accuracy and joint length recognition accuracy.Furthermore,a set of front-row driver and occupant’s body shape data visualization interface has been developed,and the height,gender,age,and weight information can be output by taking pictures of the upper body of the front-row driver and occupants in the car through binocular cameras.The research results show that this method can efficiently and accurately express the body shape information of the front-row driver and occupant in the car,which provides basic data for the development of the intelligent occupant restraint system,and has important theoretical significance and application value.