Capture Images at High Frequencies in Large-Scale Data Collection Using Augmented Mobile Mapping to Prevent Blurred Curb Signs, Despite the Increased Storage Space Required.
LADOT Pilot Tested an Augmented Mobile Mapping Technology to Digitize Curb Assets.
California, United States
Digital Curb Asset Management. Overview of LADOT’s Code the Curb Program
Summary Information
To manage curbs in cities effectively, having a comprehensive dataset is crucial in identifying existing regulations and curb assets and guiding any regulatory changes. As a part of the LADOT’s Code the Curb Program, this pilot project “Digitizing the Curb”, conducted in Summer 2021 in three Los Angeles neighborhoods and a subset of neighboring Maywood, tested an augmented mobile mapping technology for quick and scalable curb inventory utilizing machine vision to collect images of curbside signage from a car. This project digitized parking regulations and multiple assets, including bike racks, parking meters, fire hydrants, signal cabinet, traffic signal, and equipment.
The study tested whether this technology could collect curb regulations accurately and comprehensively. As the control surveying approach, manual surveying was used where pedestrian surveyors walked the curbs and documented the regulations seen via a mobile application (app). This supporting document provides more details about the pilot.
- Capture high-quality images at high frequency in large-scale data collection using augmented mobile mapping. This would prevent blurred curb signs, despite the increased storage space required.
- Implement a blended approach of augmented mobile mapping combined with pedestrian surveys to efficiently gather curb asset and regulation data and reduce costs. This study recommended that a 90 versus 10 percent blend of augmented mobile mapping versus manual surveyors, respectively, would yield effective results in digitizing curb data in cities like Los Angeles.
- Minimize driving distance by driving each street only once in each direction in the augmented mobile mapping surveying when possible. In this project, the researchers used sign detection and geolocation systems which only needed a single pass with a collection vehicle in each direction to perform well. This approach worked for low- to medium-density areas where streets can safely be driven only once in each direction.
- Optimize driving routes using parametric routing tools. Using such tools, the researchers were able to eliminate doubling the coverage of any street.
- Train machine vision models continuously for further improvements. This study revealed that training of models improves the effectiveness of machine vision applications in collecting and interpreting data.
