01836133

Monograph

https://www.trb.org/Main/Blurbs/182667.aspx

Recognizing the importance of pedestrian facility data to safety, thirty-seven State Departments of Transportation (DOTs) have prioritized improving pedestrian facility inventory, particularly concerning crosswalks and sidewalks, as an important action item in their Strategic Highway Safety Plans. However, such information is not widely available at the state level, with only 11 states having reported collection of such data. The hesitation of conducting this complex and repetitive yet essential data collection task could be raised by the challenges inherent in the current manual or semi manual data collection approaches, including human errors, high cost for time and labor, safety concerns for data collectors, and the corresponding concerns about standardizing, updating, and maintaining data. To address this urgent need for data and the challenges in the current data collection methods, this Innovations Deserving Exploratory Analysis (IDEA) project developed an innovative system using convolutional neural networks, an advanced machine learning method that uses deep learning, to process image data. These networks were used to automatically collect major pedestrian facility data, including sidewalk presence, crosswalk presence, and crosswalk length, from aerial images.

This research was conducted by the University of Southern Mississippi, Hattiesburg.

NCHRP IDEA Project 209

English

51p

2021-12

NCHRP-IDEA Program Project Final Report

Digital/other

Appendices; Figures; References; Tables

Crosswalks; Data collection; Image processing; Imagery; Machine learning; Neural networks; Pedestrian areas; Sidewalks

Innovations Deserving Exploratory Analysis (IDEA) (Program)

Data and Information Technology; Pedestrians and Bicyclists

TRIS, TRB, ATRI

Feb 18 2022 2:14PM