A method for distinction of bicycle traffic violations by detection of cyclists' behavior using multi-sensors
Masanori Ooi, Yoshitaka Nakamura, Hiroshi Inamura, Osamu Takahashi · 2016
Traffic accidents involving bicycles are the second most common kind of traffic accident in Japan. In addition, 65.1% of cyclists who had an accident had committed a traffic violation. Although various laws and regulations have been put in place, bicycle-related traffic accidents are hardly decreasing. In this paper, we propose a method for distinguishing bicycle-related traffic violations through behavior detection of cyclists using the Kinect and a photodiode. We detect cyclists using a Kinect and distinguish traffic violations of the bicycle using Skeleton Tracking, distance information, tracking time, the bicycle's travelling location, and illuminance. As experiments, we set up the Kinect and photodiode to capture the front view of a cyclist, and acted out several bicycle traffic violations outdoors. We then evaluated the distinction precision of traffic violation.