Using Computer Vision and Deep Learning Techniques to Extract Roadway Geometry from Aerial Images
The overall goal of this project is to develop computer vision tools to extract different roadway geometry data such as school zone markings, lane configurations (i.e., turning lanes lengths, and lane, shoulder and median widths), presence of signals (i.e., identification of signal poles), and sidewalks (i.e., presence or absence of sidewalks) from high resolution aerial images, which can be used by FDOT planners and engineers at various levels of traffic operations and safety analysis. Consistent with this goal, the main objectives of this project are to: (a) examine how traffic data collection can leverage emerging computer vision techniques, in particular, image processing, deep learning, machine learning, and artificial intelligence to develop statewide roadway inventory lists; (b) design an automated signalized intersection geometric data extraction algorithm based on high-resolution images in order to identify roadway geometry data such as school zone markings, lane configurations (i.e., turning lanes and their lengths), and sidewalks (i.e., presence or absence of sidewalks) from high resolution aerial images, and (c) generate a geographic information services (GIS)-based inventory list of these roadway geometry features for the entire state of Florida including ON and OFF roadways. This is an innovative solution that employs the computer vision technology to potentially replace traditional manual inventory, which is labor intensive and prone to errors.
Language
- English
Project
- Status: Completed
- Funding: $223843
-
Contract Numbers:
BED30 977-02
-
Sponsor Organizations:
Florida Department of Transportation
Research Center
605 Suwannee Street MS-30
Tallahassee, FL United States 32399-0450 -
Managing Organizations:
605 Suwannee Ste
Tallahassee, FL United States 32399 -
Project Managers:
EI-Urfali, Alan
-
Performing Organizations:
Florida State University, Tallahassee
217 Westcott Building
Tallahassee, FL United States 32306- -
Principal Investigators:
Ozguven, Eren
- Start Date: 20220330
- Expected Completion Date: 20240131
- Actual Completion Date: 20240412
Subject/Index Terms
- TRT Terms: Highway design; Image analysis; Inventory; Machine vision
- Geographic Terms: Florida
- Subject Areas: Data and Information Technology; Highways;
Filing Info
- Accession Number: 01841196
- Record Type: Research project
- Source Agency: Florida Department of Transportation
- Contract Numbers: BED30 977-02
- Files: RIP, STATEDOT
- Created Date: Mar 30 2022 3:30PM