Determination of perceived types of street crossing hazards for children under visual occlusion conditions
In order to characterize children's visual attention for street crossing under visually occluded conditions and to classify the types of hazard perceptions of children,48 children were recruited to complete the risk perception ability test experiment,and information such as children's visual attention characteristics were collected.Based on the risk perception utility quantification method,the risk perception utility values of children were quantified by children's subjective risk perception degree and situational objective risk degree.With the help of k-means clustering method,the best cluster number of 288 risk perception utility values of children in the visual field occlusion test situation was determined by elbow method,contour coefficient method and DBI index method,so the types of children's risk perception were divided into cautious type and DBI index method.The characteristics of risk perception types were selected from the aspects of children's visual attention characteristics,demographic characteristics and road environment characteristics.Linear discriminant analysis model(LDA)was constructed by machine learning and the results were compared with five common classification models,namely random forest,k nearest neighbor,naive Bayes,BP neural network and radial basis function neural network.70％of the data were used as training sets for model learning,and 30％of the data were used as test sets to test the model.The classification performance of the models was compared by macro-averages of model classification accuracy and precision,recall,and F1 score.The results show that the classification accuracy of LDA reaches 83.91％,and the model has the best generalization ability,which is most suitable for the construction of children's risk perception judgment model.The research results illustrate the group differences of children with different types of risk perception,which is helpful to understand the individual differences of children and is expected to further guide the individualized and customized development of road safety education.9 tabs,10 figs,26 refs.
traffic engineeringchildexperimental researchhazard perception typelinear dis-criminant analysisvisual field occlusion
NETL
NSTL
万方数据
