Real time dangerous driving status detection

This paper presents a method which detects the situation of car dangerous driving by video information. We estimate the parameters of global motion from the real-time video of vehicle driving, and analyze the variation of parameters for detection the dangerous status of driving. Firstly, we utilize global motion estimation to obtain the parameters of perspective model from the real-time driving video. Then, we construct the long-term and short-term model for filtering the variations of parameters. Finally, we classify the variations of parameters to define the types of dangerous driving situations. When the status of dangerous driving is detected, the alerts will try to warn the drivers to take care and send a message to the control center. Our results will be useful for assistant driving system in the future.