In Japan, the number of occupant fatalities has decreased at high ratio due to the development of passive and active safety technology in recent years. However, the pedestrian fatality is still the major proportion of fatalities according to the statistics of traffic accident. Therefore, it is important to study and develop active safety technologies which target to pedestrian. By analyzing the traffic accident statistics, narrow road in urban area is one of the dangerous parts for pedestrian. And the dangerous case that pedestrian often gets into an accident is while crossing road outside of crosswalk.
This study focuses on accidents which are caused by pedestrians suddenly crossing narrow urban roads. In such cases, drivers need to drive carefully, preparing for sudden crossing or unexpected dangerous behavior of pedestrians. Therefore, it is important for drivers to predict the accident risk due to the crossing pedestrian behavior in such narrow roads. The final goal of this study is to find out the hazard anticipation mechanism in such scenes and to design a driving assistance system, such as, a warning or intervention system. To design such a driving assistance system it is necessary to describe a model of sudden crossing or unexpected dangerous behavior of pedestrian.
To describe pedestrian behavior in urban area, near-miss incident database is analyzed which has been developed by JSAE. This database includes all kind of incidents collected by drive recorder, or black box, attached to Taxis in down town Tokyo. By using the database, we constructed a mathematical model for quantifying pedestrian behavior from camera image data, and verified the validity of the method by computer simulation.