UW WSU WSDOT




Automated Traffic Sign Recognition Using Computer Vision and Deep Learning

The importance of traffic signs for traffic operations and safety requires transportation agencies to maintain an inventory of them and their condition. To conduct such an inventory, WSDOT staff must physically visit locations for sign verification and data collection. Given the huge number of posted traffic signs, this means that traditional sign asset management is time-consuming and costly.  New, automated solutions are needed to collect traffic sign data and manage them in a timely and cost-effective manner. To address this issue, this study is developing a traffic sign data collection system from open street images, an algorithm for detecting and recognizing traffic signs in those images, and an expandable sample data inventory of traffic signs in a designated region in Washington, including both freeways and local streets. The final products will provide an automated solution to reduce manual labor and will significantly contribute to traffic sign asset management.  

Principal Investigator: Yinhai Wang, Civil and Environmental Engineering, UW
Sponsor: WSDOT
WSDOT Technical Monitor: Dina Swires 
WSDOT Project Manager: Doug Brodin 
Scheduled completion: September 2024

TRAC