Automated Video Analysis of Three Complex Brisbane Intersections

Brisbane City Council (BCC) commissioned the School of Civil Engineering and Built Environment (CEBE) at Queensland University of Technology (QUT) to assess the suitability of advanced video recognition technology in assessing road safety of transport facilities. As such, the objectives of this project were:

• To evaluate the safety performance of three complex intersections in Brisbane by applying automated video analysis techniques, and
• To translate observed patterns of conflicts to potential intersection improvement and mitigation strategies.

The intersections selected by BCC included (1) Vulture St Intersection: Vulture St—Stanley St—Dock St—Graham St, Woolloongabba, (2) McCullough St Intersection: McCullough St—SunnyPark Shops—Market Square Shops, Sunnybank, and (3) Turbot St Intersection: Turbot St—Creek St—Wharf St, Brisbane.

Automated safety analysis mainly includes three components: automated volume count, conflict analysis, and violation analysis. Automated volume count involves the count of various types of road users, e.g. passenger cars and pedestrians. Conflict analysis includes counting conflict frequency, and identifying conflict severity and location (conflict points). The conflicts observed at an intersection mainly include vehicle-vehicle interactions, vehicle-pedestrian interactions and vehicle-cyclist interactions. Violation analysis includes the automated identification of non-conforming behaviour of road users. In this project, two types of conflict indicators were considered: Time-To-Collision (TTC) and Post Encroachment Time (PET).

This project demonstrates the capability of automated video analysis technique in diagnosing safety of complex intersections and identifying targeted countermeasures. In addition to assessing safety of other transport facilities, this safety assessment tool can be applied to investigate particular road user behaviour such as manoeuvres of cyclists at roundabouts, lane filtering of motorcyclists, and jaywalking behaviour of pedestrians.