Faculty Sponsor
Dr. Thobias Sando, PhD., P.E., PTOE.
Faculty Sponsor College
College of Computing, Engineering & Construction
Faculty Sponsor Department
Engineering
Location
SOARS Virtual Conference
Presentation Website
https://unfsoars.domains.unf.edu/automated-extraction-of-crosswalk-from-high-resolution-images/
Keywords
SOARS (Conference) (2020 : University of North Florida) -- Posters; University of North Florida. Office of Undergraduate Research; University of North Florida. Graduate School; College students – Research -- Florida – Jacksonville -- Posters; University of North Florida – Graduate students – Research -- Posters; University of North Florida. Department of Engineering -- Research -- Posters; Engineering; Math; and Computer Sciences -- Research – Posters
Abstract
Intersection geometric data are crucial in addressing various safety and operation issues in roadway intersections. Traditionally, agencies have hire staff to extract geometric information from images visually and manually. Manual data extraction is time-consuming, costly (labor intensive), and prone to errors. This study aims to develop an automated signalized intersection geometric data extraction algorithm based on high-resolution images in order to identify crosswalks, left and right turns, and mid-block crosswalks. Using the computer vision technology, the developed algorithm will be able to extract key intersection geometrics such as intersection width, number of lanes, intersection configuration, the presence of turn lanes, pedestrian crossing (crosswalks and mid-block crosswalks), median type, and other pertinent intersection features that can be recognized from images. This presentation discusses the framework of the automatic intersection data extraction process starting with the data collection, processing, analysis and evaluation of the results. The presentation also outlines some of the challenges associated with automation and the recommended methods for overcoming such challenges.
Included in
Automated Extraction of Crosswalk from High-resolution Images
SOARS Virtual Conference
Intersection geometric data are crucial in addressing various safety and operation issues in roadway intersections. Traditionally, agencies have hire staff to extract geometric information from images visually and manually. Manual data extraction is time-consuming, costly (labor intensive), and prone to errors. This study aims to develop an automated signalized intersection geometric data extraction algorithm based on high-resolution images in order to identify crosswalks, left and right turns, and mid-block crosswalks. Using the computer vision technology, the developed algorithm will be able to extract key intersection geometrics such as intersection width, number of lanes, intersection configuration, the presence of turn lanes, pedestrian crossing (crosswalks and mid-block crosswalks), median type, and other pertinent intersection features that can be recognized from images. This presentation discusses the framework of the automatic intersection data extraction process starting with the data collection, processing, analysis and evaluation of the results. The presentation also outlines some of the challenges associated with automation and the recommended methods for overcoming such challenges.
https://digitalcommons.unf.edu/soars/2020/spring_2020/67