Study on driver's unsafe gaze behavior evaluation technology
Wang Pen · Zhongguo anquan kexue xuebao · 2014
For improving safety of drivers' gaze behavior,a driver unsafe gaze behavior recognition system was established by vision detection technology. In this system,Characteristic information T was made up of interpupillary distance, and distance between mouth and midpoint of two eyes. The interpupillary distance was used to measure the behavior of looking right and left. The distance between mouth and midpoint of two eyes was used to measure the behavior of looking up and down. The characteristic information T was combined with the driver's eyelid closure and proportion and location characteristics of iris-sclera in order to characterize driver gaze status. Combing SVM,multi-information decision level fusion was proposed to estimate the different modes of driver's gaze behavior. The simulation results indicate that driver unsafe gaze behavior can be evaluated on basis of multi-source information fusion.