Name
Abstract | ||
|---|---|---|
In most of the traffic safety studies, both the identification of high-risk locations and the assessment of safety improvement solutions are done through the use of historical crash data. This study proposes an alternative approach that makes use of traffic conflicts extracted from traffic video recordings for safety assessment. State-of-the-art computer vision techniques are used to extract vehicle trajectories automatically from 70 hours of traffic video data at two intersections. More specifically, a modified implementation of the Kanade-Lucas-Tomasi (KLT) feature tracker is used to extract the feature points and track those feature points frame by frame. The spectral embedding and the Dirichlet process Gaussian mixture model (DPGMM) are employed to cluster feature points that belong to the same object. The combination of each vehicle's individual trajectory with all the others' trajectories is then screened to identify all the possible vehicle pairs involved in conflict risk. Traffic conflict risks are identified after the time to collision (TTC) is computed for each vehicle pair. Hourly number of conflicts are found to follow a negative binomial distribution similar to number of crashes. A strong correlation is observed between the traffic conflicts and actual crashes, and thus the validity of using conflict data extracted from videos for safety assessment can be confirmed. The proposed approach has potential transferability and can be implemented by transportation agencies in other cities. |
Year | Venue | Field |
|---|---|---|
2016 | 2016 IEEE 19TH INTERNATIONAL CONFERENCE ON INTELLIGENT TRANSPORTATION SYSTEMS (ITSC) | Computer vision,Crash,Embedding,Simulation,Feature extraction,Artificial intelligence,Negative binomial distribution,Time to collision,Engineering,Transferability,Trajectory,Traffic conflict |
DocType | Citations | PageRank |
Conference | 0 | 0.34 |
References | Authors | |
0 | 7 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Kun Xie | 1 | 3 | 1.38 |
| Chenge Li | 2 | 4 | 1.10 |
| kaan ozbay | 3 | 26 | 12.33 |
| Gregory Dobler | 4 | 1 | 1.34 |
| Hong Yang | 5 | 22 | 3.74 |
| An-Ti Chiang | 6 | 0 | 0.34 |
| Masoud Ghandehari | 7 | 1 | 2.69 |