home   /   centre for automotive safety research   /   Publications   /   List   /   Details

Publication Details

TitleVideo capture and analysis of cyclists using infrastructure in the ACT through machine learning
AuthorsMackenzie JRR, Ponte G
AbstractThe implementation of infrastructure to reduce traffic conflicts and improve road safety for cyclists is critical. However, potential benefits resulting from strategic interventions can only be achieved if there is corresponding compliance or good utilisation of that infrastructure. Traditional methods for evaluating cycling behaviours and interactions with the traffic system either involve expensive roadside observations, which can influence cycling behaviours through the observer effect, or by using induction or pneumatic tube system which can only yield count information and cannot be used in mixed traffic situations. This study used long duration portable video cameras to, somewhat covertly, record mixed road traffic, with the captured video subsequently being processed with bespoke machine learning software. This innovative process required negligible set-up or manual processing time and a researcher was only required to do a desktop analysis on 6% of the total traffic video recorded. Agencies requiring evaluations of cyclist interactions with specific infrastructure or behaviours at specific locations could use this rapid and cost effective process to assist with data collection and analyses to inform or optimise their cycling safety strategies.
Report NumberCASR202
PublisherCentre for Automotive Safety Research
Publisher CityAdelaide
SponsorThis project was made possible with assistance from the ACT Road Safety Fund
Page Count39
NotesReport is freely available on funding organisation website also >>

Mackenzie JRR, Ponte G (2022). Video capture and analysis of cyclists using infrastructure in the ACT through machine learning (CASR202). Adelaide: Centre for Automotive Safety Research.

Files Available for Download
CASR202.pdfPDF of final report