Computer vision-based model for detecting turning lane features on Florida’s public roadways from aerial images

ABSTRACT

Efficient collection of roadway geometry data is crucial for effective transportation planning, maintenance, and design. Current methods involve land-based techniques like field inventory and aerial-based methods such as satellite imagery. However, land-based approaches are labor-intensive and costly, prompting the need for more efficient methodologies. Consequently, there exists a pressing need to develop more efficient methodologies for acquiring this data promptly, safely, and economically. This study proposes a computer vision-based approach to detect turning lane markings from aerial images in Florida. The method aims to identify aged or faded markings, compare lane locations with other features, and analyze intersection crashes. Validation in Leon County achieved 80.4% accuracy, detecting over 13,800 turning lane features in Duval County, Florida. This data integration offers valuable insights for policymakers and road users, highlighting the significance of automated extraction methods in transportation planning and safety.

KEYWORDS:

• Turning lanes
• deep learning
• roadway characteristic index (RCI)
• pavement markings
• machine learning (ML)
• roadway geometry features

Acknowledgements

The following authors confirm contribution to the paper with regards to Study conception and design: Richard Boadu Antwi, Samuel Takyi, Kimollo Michael, Alican Karaer, Eren Erman Ozguven, Ren Moses, Maxim A. Dulebenets, and Thobias Sando; Data collection: Richard Boadu Antwi, Samuel Takyi, Kimollo Michael, Eren Erman Ozguven, Ren Moses, Maxim A. Dulebenets, and Thobias Sando; Analysis and interpretation of results; Manuscript preparation: Richard Boadu Antwi, Eren Erman Ozguven, Ren Moses, Maxim A. Dulebenets, and Thobias Sando. All authors reviewed the results and approved the final version of the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was sponsored by the State of Florida Department of Transportation (DOT) [grant number BED30-977-02]. The contents of this paper and discussion represent the authors’ opinions and do not reflect the official views of the Florida Department of Transportation.