Program Title

Identifying Roadside Hazards
MR PR3 The Detail Dilemma

MR Problem

Download PDF

How might we better

Identify roadside hazards from video footage

Problem Story

There are thousands of road assets spread across the 18000kms rural road network in WA. Main Roads updates the risk associated with sections of road to identify future areas for improvement.
We have video footage of our network. Can we use this to identify hazards on our road to enable us to quantify risk?
Examples of hazards include trees, rocks, power poles, crash barriers, etc). Developing techniques to identify hazards from video, locate them and estimate their size and distance from the road would streamline the updating of our risk ratings.

Scale / Impact


Video data already captured

Potential hazard to road users if assets misidentified or not recorded

Cost to manually review thousands of hours of video

Data Description

Main Roads Video surveys
Laser/pointcloud data

External Satellite data?

4 key outcomes / constraints / statements

  1. 18,000km of rural road network
  2. Existing video footage of the entire network
  3. Sample laser/pointcloud data from some of the network
  4. Automation of hazard detection using images