Study on a Selection Method of Objects contribute to Driver Operation based on a Statistical Driving Behavior Model

Abstract

In order to assist recognition of driver, it is effective for driver to teach recognition objects in a visual way. However, it is expected that too much information is provided for a driver from the system and it cause him distraction. Therefore, information presentation without exaggeration and without omission is demanded for assistant system. In this paper, it is assumed that the objects which contribute to driver’s braking operation should be presented to him. However, these objects changes according to the facing driving situation. Therefore, a selection method of these objects in the appeared objects on driving environment based on a statistical driving behavior model is proposed in this paper. In this method, a driving behavior model is generated, which is consisted of objects detection model with deep neural network structure and time series correlation model between the appeared objects and braking operation with probabilistic model structure. The probability of contributing to braking operation for all appeared objects in driving environment is calculated based on the driving behavior model, and the objects with the high probability are selected as the object which contributes to braking operation.In the experiment, the selection and presentation accuracy of the object which contributes to braking operation was examined. As the results, it was confirmed that the appropriate object can be selected by using the proposed method, and this method has an effect of reducing false or unnecessary presentation information.