Abstract:

An approach about dangerous driving behavior recognition was proposed to solve the problem of uncertainty and complex in driver's behavior identification. This approach was based on a combination of Dempster-Shafer (D-S) evidence theory with entropy weight grey incidence considering both the comprehension
of different indexes and the uncertainty of different goals. The weights of indexes were calculated by entropy theory to determine uncertainty reliabilities with grey relation analysis and to construct a Mass Functions for different goals. Dangerous driver status was identified based on Dempster synthesis rule with D-S theory evidence that integrated Mass functions. An expert evaluation method based on facial video was used to judge driving behaviors. The experimental results show that the method provides the recognition accuracy of 91.25% under high-speed condition. Therefore, the reliability and accuracy of the identification method are significantly higher than those of single sensor are.

Key words: vehicle safety , dangerous state of fatigue driving ,  recognition methods , Dempster- Shafer (D-S) evidence theory ,  entropy weight