In accordance with the CoA Vision Zero Action Plan, the City of Austin has an ambitious goal of reducing traffic crash deaths and serious injuries. CTR and the Texas Advanced Computing Center (TACC) have developed a method to use automated object recognition/tracking in video data streams generated from traffic monitoring cameras for applications that may support the Vision Zero initiative.
Potential uses of the developed methods include generating insights on traffic and pedestrian patterns, recognizing abnormal traffic behavior, assessing the impact of pedestrian safety measures, and studying other safety concerns, such as near-misses.
The tools and exploratory work conducted in previous years will be refined to enable the following applications:
- Compare/characterize pedestrian patterns before/after the deployment of PHBs at selected locations where visibility is adequate.
- Compare/characterize pedestrian activity across selected locations to support the prioritization of improvements.
Three types of refinements will be considered:
- Developing and implementing metrics to visualize/quantify the results from automated analysis in order to support decision making and system safety evaluation.
- Extending algorithms to allow for more precise pedestrian counting (exploratory).
- Developing a web interface to facilitate access and querying of video data files, including features to facilitate manual data analysis (e.g. identify all frames where pedestrians are present in one or more files, or where pedestrians and vehicle interactions are expected).
Read the final report here.