| Title | Road Incident/Accident Detection and Image Analysis |
| Dept | Transportation Safety Division |
| Year | 2009 |
| Month | |
| Price | |
| Summary | Accident numbers, which can only be acquired from accident reports made by police officers, are important indexes for accident prone location selection and countermeasure improvement evaluation. There are many reasons for why accident reports can’t describe the accident facts properly. First of all, police officers always finish this report after the accident happened and it’s hard to collect evidence. Second, the most frequently used evidence is the witness testimony which may be biased after the event. Third, we don’t know how accident affects the upstream traffic flow. Finally, we presume or judge the accident cause based on static data instead of dynamic data, i.e. the accident process. Therefore, if we have real-time video data with the help of an incident detection system for accident black spots, we can not only shorten the notification time but also improve the effects of countermeasures for accident prone locations. We selected two accident prone locations in Taipei, including one intersection and one mid-block of arterial and record video continuously for a whole month at each site. We firstly used these videos to test Citilog, a French incident detection system, in urban areas. Secondly, we analyzed the accident video with our own traffic flow image analysis tool to get traffic parameters, like q, k and v, for further discussion. At the intersection, four accidents were recorded and found that under-reporting exists when compared with the police accident reports. We also analyzed the impact of accidents on upstream traffic flow then made some suggestions for countermeasures. No accidents in the mid-block location were recorded, so we analyzed the impact on upstream flow when there was an incident on roads instead. Furthermore, this was the first attempt to use a video incident detection system in an urban intersection and we accumulated very precious experience from this project. We kept these experiences in an appendix for future reference. |
| Post date | 2009/12/02 |
| Hashtags |
View count:
115