Traffic sign detection and positioning based on monocular camera
Journal
Journal of the Chinese Institute of Engineers, Transactions of the Chinese Institute of Engineers, Series A
Date Issued
2019
Author(s)
Abstract
Geospatial and geometric states of traffic signs are indispensable information for traffic facility maintenance. This study applies a low-cost single camera system to locate and identify traffic signs and reflects their on-site conditions for assisting in maintenance. Referring to official regulations, a traffic sign can be identified and classified based on its color and shape attributes. The poses and states of the sign planes can also be evaluated according to their geometric shape ratios and the discrepancy with respect to traffic sign regulations, in which a least-square consistency check is proposed to ensure assessment reliability and accuracy. Validation with day and nighttime image sequences was performed to verify the effectiveness of the proposed method and the robustness to illumination changes, color deterioration, and motion blur. Considering the promising results, this study can be deemed as an alternative to promote routine maintenance of traffic facilities.
Type
journal article