Analysis Of Driving Behavior Based On Dbn And Its Application

With the rapid development of society and economy,the number of automobiles is increasing.How to effectively prevent road safety accidents has become the focus of traffic research in recent years.Driver's driving behavior is the most important factor affecting traffic safety.Focusing on and studying driver's behavior is of great significance to guard against traffic crash,accelerate driving auxiliary system developing and fleet management to evaluate driver's behavior.Based on the driver's behavior data collected by the on-board system,this thesis uses deep belief network algorithm and comprehensive evaluation method to mine and analyze the stored data,identifies the driving style,proposes a comprehensive evaluation system for drivers based on four criteria: driving safety behavior,driving fuel consumption,driving experience of unequalled pleasure and driving skilled,and comprehensively uses the processed data to drive.The primary coverage of the thesis are listed next:(1)In order to identify driving behavior style,this study proposes a DBN driving behavior style recognition model based on multi-layer supervised training and learning fine-tuning.Firstly,this thesis introduces the knowledge of driving style,analyses its influencing factors,extracts the corresponding features,and uses K-means to cluster the feature data.At the same time,a DBN recognition model with multi-layer supervised training and fine-tuning learning is established,and the clustering results are used as input of the model to recognize driving behavior style.Compared with the classification result from primitive DBN and BP model,the tests prove that the recognition results of the model in this thesis is better than that of the original DBN model and BP model.(2)In view of the lack of a comprehensive evaluation system for driving behavior habits at home and abroad,this study proposes a comprehensive evaluation system model of driving behavior founded on the community of analytic hierarchy process and entropy method.Considering the low cost of GPS input and the advantages of easy data collection,this thesis first proposes a multi-criteria comprehensive evaluation system of driving behavior,including four criteria: driving safety,driving fuel consumption,driving experience of unequalled pleasure and driving skilled.Afterwards,the thesis uses the combination method of AHP and entropy method to calculate the weight of each index in the model,and then makes a quantitative analysis of the driver's criteria and indicators,builds a comprehensive scoring model of driving behavior,and analyzes the specific driver's example.(3)On the basis of the above research,this thesis will design and implement a system of driving behavior analysis and its application.In the system,the results of comprehensive evaluation of driving behavior style and driving behavior are visualized,and system functions such as bad driving statistics,driving trajectory playback and other data visualization display are realized.