The (Not So) Secret Formula for Citywide Digital Curbs

Jake Belman
January 18, 2022

The size of Los Angeles County is truly mind-boggling. At just over 4,000 square miles, it is larger than Rhode Island and Delaware combined and home to more than ten million people. As for streets, Los Angeles proper has about 14,000 miles of roadway.

In 2016, the Los Angeles Department of Transportation (LADOT) launched Code the Curb, an initiative to create a digital inventory of the county’s curb regulations. This would require surveying the curb regulations on the ground: like most cities in North America, Los Angeles’ curb regulations are not documented. There are tools to make this surveying easier, but all of them are for surveying by foot.

Let’s do some quick math. We’ve found that surveyors can cover about one mile of curb by foot per hour. Each mile of street equals two miles of curb, so covering Los Angeles’ 14,000 miles of street by foot would take the average person 28,000 hours. For ten surveyors working 40 hours a week, 48 weeks a year, this would take about a year and a half – and probably cost close to a million in salaries.

Put simply, this is not a scalable solution.

A Faster Method for Curb Inventories?

What’s faster than walking? Cycling Driving, of course.

Get in loser, we're going surveying. Picture of a scene from Mean Girls.

Surveying while driving has a name: mobile mapping. It’s when you use a camera mounted on a car to map streets, assets, etc. while driving. We took this process a step further by building Curb Converter, a machine vision architecture that augments the mobile mapping process by finding curb signs in each image and converting them into digital curb data. We call it augmented mobile mapping.

Augmented mobile mapping has the benefit of being a very efficient way to collect data, but there are some tradeoffs: GPS signals can be blocked by tall buildings, signs might be obstructed when you drive past them, and image quality varies. Over at CurbIQ headquarters, we wanted to find out if we could overcome these issues without breaking the bank. So we partnered up with Urban Movement Labs (UML) in Los Angeles to test it out.

First, we picked four neighbourhoods in Los Angeles that differed in terms of their use patterns and density. Then, using only smart phone cameras and GPSs, we got to work driving the streets and collecting data.

The detailed results of this pilot, and our takeaways, can be found in our final report Digitizing the Curb: Curb Inventory Pilot Project .

How Well Does Augmented Mobile Mapping Work

When all the images were uploaded, Curb Converter had done its magic, and all the regulation data was processed, we got our answer.

In high-density areas, augmented mobile mapping works quite poorly. Here are the results for the highest-density neighbourhood we surveyed, South Park in DTLA.

High Density Neighbourhood Survey Results:

  • Detection: 93% of curb regulation signs were detected
  • Comprehension: 73% of detected signs were perfectly read and understood
  • Geolocation: 91% of detected signs were mapped by our system within two car lengths of their actual locations, and only 57% were mapped within one car length

These results may not look too bad, but nearly 1 in 10 signs missing from a digital curb inventory is abysmal considering the impact on the data’s end-users. And less than three quarters of the signs that were detected were even accurately interpreted by Curb Converter. Overall, that means that only 68% of signs were detected and comprehended perfectly.

Grade: B-

The other side of the coin is that augmented mobile mapping is very effective in other parts of the city. In low- to mid-density neighbourhoods, including practically all residential areas, augmented mobile mapping works just about as well as surveying by foot and is up to eight times faster. Here are the results in Maywood, a moderately dense residential neighbourhood and the lowest-density area we surveyed:

Low/Moderate Density Neighbourhood Survey Results:

  • Detection: 98% of curb regulation signs were detected
  • Comprehension: 100% of detected signs were fully comprehended
  • Geolocation: 98% of detected signs were positioned within two car lengths of their actual locations, and 82% were positioned within one car length

That’s pretty good! Nearly 100% of signs were detected and perfectly comprehended.

Grade: A+

CurbIQ’s Secret Formula for Scalable Curb Inventories

Okay, so augmented mobile mapping is great in low- and moderate-density neighbourhoods and not viable in high-density neighbourhoods.

Fortunately, low- and medium-density neighbourhoods like Maywood make up a huge proportion of North American cities. In Los Angeles, we estimate that around 90% of the curb miles in the county are use-types very similar to Maywood, where augmented mobile mapping works extremely well. If we can survey 90% of Los Angeles eight times faster than we ordinarily could have, we’re still surveying the entire city in just over a fifth the time it would take to survey it by foot alone.

So here’s the secret formula for scalable curb inventory:

  • Mobile map in low- and medium-density areas; use computer vision to do the heavy lifting
  • Survey by foot in high-density areas, like dense commercial corridors

What Does This Mean?

Curbside data is super valuable for cities and the folks who run them. Here are just a few things cities can do with citywide curbside data:

  • Share an interactive parking map with citizens so drivers can head directly to legal parking zones rather than meandering around (all the while creating congestion and emissions!)
  • Recover new revenue by identifying neighbourhoods with an excess of free parking and making some of it paid
  • Evaluate the impact of large-scale street redesign projects on parking revenue, delivery access, and accessible parking with real data
  • Identify candidate segments for new curb uses, like micromobility docking stations, pick-up/drop-off zones, and electric vehicle charging stations
  • Empower entrepreneurs in your city to develop apps and software that help drivers, cyclists, delivery companies, etc. to make the most of the curb

(By the way, curbside data has a lot of benefits for engaged citizens too.)

Maybe you’ve been eyeing a new BRT line in your city but are having a hard time financially justifying it. Maybe you’ve been tasked with helping your city to meet its sustainability goals. Or maybe your constituents are telling you that parking is too expensive in your district, and you want to look at the data yourself.

Whatever your challenge, if you think that citywide curbside data could help you or want to find out if it can, give us a shout. We’d be glad to help you through it.

If you’re interesting in learning more about citywide curbside data, watch the recording of our recent Digitizing the Curb webinar.

Jake Belman
Jake is a Transit Technology and Data Analyst who specializes in the design and development of custom intelligent transportation system solutions. He is an experienced computer programmer and received his bachelor's degree in Mechanical Engineering from McGill University.
Global trends on curbside management
Curbside solutions delivered to your inbox, 1x per month, for free.
© Copyright - CurbIQ | Privacy