Research On Aggressive Driving Behavior Identification Based On Multi-source Heterogeneous Data Fusion

Aggressive driving refers to driving behaviors that cause psychological or physical harm to others due to anger or irritability.In terms of traffic behavior,it includes frequent lane changes,preventing other drivers from overtaking,passing red lights,speeding,and so on.Most of the researches on aggressive driving behaviors focus on the macro analysis and static discrimination of the driver's aggressive driving behaviors.How to quantify and identify the aggressive driving behaviors has been a difficult research point.Driving behavior is usually divided into macro driving behavior and micro driving behavior.Macro driving behavior is classified according to the causes and results of behavior,such as fatigue driving,drinking driving.Aggressive driving behavior is also a driving behavior in a macro sense.Micro driving behavior mainly refers to the external performance of the driver's specific operations,such as car following,overtaking,lane changing.This article analyzes and studies the aggressive driving behaviors in a macro sense from a micro perspective.Based on the multi-source data obtained in the experiment,combined with data fusion theory to analyze the aggressive driving behavior,and then quantitatively identify the aggressive driving behavior.This is of great significance for the advancement of personalized driving safety warnings.Firstly,a multi-source information acquisition system based on the Psy LAB human factors engineering experimental system and a multi-person interactive driving simulator was constructed to obtain driver driving behavior data under the aggressive induction mechanism.Secondly,the internal structure of the aggressive driving behavior is analyzed from the horizontal and vertical perspectives.What kind of basic abnormal driving behavior does the aggressive driving behavior consist of and extracts the abnormal driving behavior fragments.The impact of micro abnormal driving behavior on aggressive driving behavior is analyzed.Thirdly,these abnormal driving behavior fragments use the average influence value and artificial neural network to extract feature parameters,and use BP-Adaboost algorithm to identify abnormal driving behavior.Finally,a generalized neural network is used to determine the clustering center and the fuzzy c-means clustering model is optimized.Use the number of times of abnormal driving behavior to identify normal driving behavior,first-level aggressive driving behavior,and second-level aggressive driving behavior.The results show that the established quantitative identification model of aggressive driving behavior can effectively distinguish between normal driving behavior,first-level aggressive driving behavior and second-level aggressive driving behavior.The research in this article enriches the research on aggressive driving behaviors,contributed to the quantification and identification of aggressive driving behavior,provides theoretical support for driving safety warning.