Automatic detection method research of incidents in China-FOT database

Field Operational Tests (FOT) or Naturalistic Driving Studies (NDS) is necessary to evaluate advanced driver assistance system (ADAS) as well as the naturalistic driver behavior. Due to incidents (related to crash events) may lead to collision accidents, it is necessary to do the related studies on incidents to improve traffic safety. In these studies a large volume of data was collected and a lot of time had to be spent to extract dangerous incidents and therefore the efficiency was extremely low. An automated method to detect dangerous incidents according to the characteristics of Chinese traffic and driver behavior was proposed in this paper to filter the China-FOT database. In this method false negative rate of this method was firstly reduced and false positive rate was then reduced by a fuzzy comprehensive evaluation method based on vehicle dynamics parameters. The efficiency of detecting dangerous incidents was greatly improved by this method.