Deep Dive: Methods and considerations for collecting curb regulation data
Increased interest in the “humble curb” has prompted more and more cities to consider inventorying their curbspace. But how can they go about doing this? What methods and tools are available? And what exactly needs to be collected?
No matter how a city tackles a curb inventory, imagery and geolocations need to be collected from the field — with dashcams or LIDAR, or by staff on the street. These raw inputs must be processed (by humans or computers) into structured curb regulations with street-referenced geometries. Cities can use existing, free tools (like FieldPapers + a camera + JOSM), to do this fieldwork and digitization themselves. Or they may want to work with a vendor with more sophisticated tools.
This post outlines some of the main considerations cities have when collecting this data, such as location accuracy, standardized outputs, and ownership of the processed data. We offer thoughts on the task itself and discuss different methods for collecting the data.
Objectives of a curb inventory
There are many ways to capture curb regulation information, depending on whether a city prefers to leverage technology or labor. Some cities hire vendors with sophisticated imagery-processing algorithms. Others send out an intern with a notebook and a pencil. Some do both, as a way to verify the data. Regardless of which path a city chooses, it’s important to first understand what’s actually involved in a curb inventory and what types of data will be needed.
Inventorying curbspace involves mapping two different things: the physical assets located along the street, and the legal regulations that govern sections of the street. Physical assets include parking meters, signposts, curb paint, fire hydrants, and other infrastructure that communicate rules to the public. Regulations are invisible, temporal usage restrictions that govern which uses of the curb are permitted, when, and by whom, along a particular segment of the street. Assets and regulations are two different concepts, each with its own geography, like so:
Cities need data about both the assets and regulations on a street. Therefore, a curb inventory needs to accomplish three things:
- Capture the essential asset information
- Convert the assets into street sections
- Store usage restrictions in a structured, readable format
This is true no matter how the data are actually collected, or by whom. When hiring a vendor or undertaking this work internally, a city should ensure that the curb inventory (and any associated deliverables) will achieve these three objectives.
1. Capture essential asset information
Field data collection typically involves collecting the minimum viable information on the ground and then processing this afterwards. Based on what a curb inventory needs to accomplish, here’s the minimum information that should be collected on the street:
Photograph(s) of each curb asset
Best practice is to collect one of more photos of each curb asset. Parking signs and meters often communicate complicated information and it’s generally much faster, easier, and less error prone to process this information afterwards. Taking photos creates imagery that algorithms can process or humans can digitize.
Geographic location(s) of each curb asset
Each signpost, parking meter, fire hydrant, and other curb asset needs to be mapped. These assets will need to be converted into street segments, so it’s best to ensure that each restriction has a start and end point. This means, for example, marking both the start and end of a crosswalk, or the start and end of where a fire hydrant restricts parking.
There are several ways to determine locations. Cell phones and other devices can determine their geographic coordinates using GPS, wifi triangulation, and/or cell tower triangulation. However, location accuracy limits and urban canyoning can make it difficult to reliably generate precise coordinates. For more accurate data, surveyors could note the location on a map where each photo was taken. For the most accurate data, surveyors could use a measuring wheel to note where each photograph was taken.
More automated curb inventories often involve video or image processing. To generate accurate location information for each curb asset, these methods must be able to determine the exact location where the image was taken, as well as the exact location of the sign that’s depicted in the image. This is a difficult problem, so many images from a location are used as an input; this allows for a high number of points to triangulate from and results in greater accuracy. When using this approach, location accuracy and confidence intervals should be discussed with potential vendors so that a city knows what sort of precision it can expect.
3. Optional: Angle of parking
There are some types of street data that are useful for a curb inventory, but which cannot be determined from sign photographs or geographic coordinates. For example, a city may want to record the angle of parking (parallel, perpendicular, or angled) on each street in order to estimate the number of parking spaces governed by each regulation. It could also record the exact number of spaces in each regulation, if that data were needed. This type of information could be quickly recorded during field data collection. Or a city could consult satellite or street-view imagery during data processing.
2. Convert the assets into street sections
Once the point-based asset data is processed, it should be converted so that it references a given street, at given particular length. For example, a set of signposts may indicate a “Resident Permit Parking” zone along Elm Street, from 10 meters to 50 meters.
One option for this would be to add this information into a city’s street centerline GIS data or a company’s internal map. This approach is messy, however — a street often has four or more overlapping curb regulations on each side of the street, so splitting road geometries quickly becomes messy. Moreover, this approach locks data into particular representations of the street. If a neighboring city uses a different basemap, then there isn’t a way to communicate about curb regulations in a way that everyone can interpret.
A better method is to use linear referencing, which provides a way for cities to reference a given street, side of the street, and location along it. Cities like DC use this approach internally, but this approach requires each city to build a linear referencing model for themselves. As an alternative, the SharedStreets Referencing System provides an open, non-proprietary, seamless way to reference streets for any city in the world. Existing tools help facilitate the conversion of asset data into street segments. Cities can input a set of curb asset points and the tool will snap these to the nearest street and establish a length using either relationships between signs, or a prescribed buffer width for each feature. The output is linear-referenced street segments that can be used in standard GIS software.
When collecting data or working with a vendor, cities should be explicit about how these regulatory geometries will be created so that the final output meets their needs.
3. Store usage restrictions in a structured, readable format
Structured data about the curb usage rules needs to be attached to both the asset points and the regulatory geometries. Cities should consider who is doing this digitization (a vendor, or city staff?), the process for digitization, and what the outputs will be.
The output format is particularly important. Let’s say you’re using sign photographs to build a GIS file for parking regulations. What fields do you include? What’s the correct way to describe a taxi stand? How would you store data for a street that is a commercial loading zone during rush hour, then a two-hour metered parking zone at off-peak hours?
Until recently, there wasn’t a standardized way to store and share curb regulation information, which meant that each city had to navigate this problem on its own and it was harder for cities to be explicit about what they needed from a vendor.
Earlier this summer, SharedStreets launched CurbLR, an open curb data standard that cities and companies can use to communicate about the curb in a structured format. We worked closely with several cities in order to develop the standard, in partnership with Ford Mobility’s UK-based team.
The standard provides a way to share curb information with companies and leverage visualization and analysis tools across different jurisdictions. It also provides a target for data collection and vendor deliverables. By shaping their data collection and processing around what’s needed to generate a CurbLR feed, cities can ensure that they are collecting all the essential data the first time around. By requiring a vendor to deliver data as a CurbLR feed, cities know that the final output will include street-linked geometries that contain essential, standardized data. This ensures that cities get what they need and prevents vendor lock-in if a city wishes to take their data to a different vendor, platform, or service. This gives cities greater ownership over their data and processes.
For these reasons, we highly recommend CurbLR as one of the target outcomes for any curb inventory.
Methods and tools
So, what are some of the options for creating standardized curb data?
High-tech option: Imagery processing
Many private companies will collect and/or process street-level imagery in order to extract curb regulations. This method is often used by cities who need to map large areas, or who require regular updates.
Advantages: This method requires less (or no) staff time spent doing field data collection. It may or may not involve data processing by city staff, depending on the vendor outputs. Assuming a technology works well, imagery can be collected, processed, and updated fairly quickly.
Considerations: Cities should consider the location accuracy of potential vendors. They should discuss what data deliverables will be to ensure that they receive both the asset data and the regulatory geometries in an appropriate format, along with standardized regulation information (again, we recommend requiring a CurbLR feed as a deliverable). Finally, cities should consider data ownership. Some vendors maintain control over the resulting data; cities pay a regular subscription fee for this data, while companies maintain the right to sell the data to private companies. If a city is paying for data collection, we recommend that they own the rights to the output so they can use it indefinitely, port it to other vendors or platforms if desired, and disseminate publicly if they wish.
Low-tech option: Field data collection tools
Various mapping and data collection tools already exist and could be used for curb surveying. These methods involve sending surveyors on the street to manually collect location information and photos, then process the data. This approach may suit cities who are looking to collect data for a smaller area, who want a greater degree of control over data collection and processing, or who would prefer to use manpower over AI.
We field-tested two Android apps designed for structured field data collection and mapping: OpenDataKit (also available as KoBoCollect) and OpenMapKit. For curb surveying, we weren’t satisfied with either because of the map interfaces; it was difficult and time-consuming to collect a precise location for a point of interest while standing on the street. In the case of OpenDataKit and KoBoCollect, only the current asset point could be viewed, which made it impossible to place asset points relative to one another.
Our preferred method was Field Papers and a camera. Field Papers enables users to create a paper-based map atlas for any area of the world, at a desired zoom level. Users annotate the maps in the field and then scan and upload them back at the office. Each map page is georeferenced automatically and can be used to create map-based information.
Field Papers was designed for the creation of OpenStreetMap data, but it was easy to repurpose it for curb surveying. We printed the atlas pages for a nearby commercial district and headed out to survey it. On the street, we took a photo of each curb asset and noted the asset’s precise location on the map, along with a code to indicate the type of associated regulation (e.g. “H” for “handicap permit”). Since the atlas pages were at a high resolution and included business names and landmarks, it was easy to note a fairly accurate location for where each asset was located — these are the numbers and shorthand visible in the image below.
After capturing photos and annotations for the area, we headed back to the office to upload the atlas pages and digitize the data.
Digitization takes place in JOSM, an OpenStreetMap editor. In the background, we have the annotated, georeferenced atlas page. On top of that, we have the asset photos captured on street — these are visible as tiny camera icons in the image below. Each photo has a GPS location and appears close to the associated map annotation, but not exactly, due to location accuracy limits.
The purpose of digitization is to capture the location of the map annotations and attach the content of the asset photo, in a structured format. To do this, we added a point feature (visible as a red dot) on top of each asset location noted on the map. Then, we digitized the curb rule using the photographs, adding tags to each point based on the CurbLR standard. Shorthand notes on the map made it easy to add these tags for multiple points at a time (e.g. adding tags for all handicap signs at once).
While this data could be uploaded to OpenStreetMap (see our previous discussion on this), the purpose of this exercise was to create internal data and a CurbLR feed for a city, so we did not upload it. Instead, the data was saved as a
.osm file and converted into a
.geojson, which can be read by GIS applications. We used the SharedStreets matching tools to convert the points into linear-referenced street segments, then converted the file into a CurbLR feed.
Advantages: This method is free, open source, and available for cities to try out. The city fully owns the process and the resulting data. Field data collection is straightforward and can be done by surveyors without a GIS background or specialized skills. Reasonably accurate location data can be gathered through the map-based annotations, and surveyors could also incorporate a measuring wheel if extremely precise locations were needed.
Considerations: This approach requires more time and labor than more technology-intensive methods like imagery processing. Some technical skills are needed for data processing and conversion. While this approach does meet the requirements, JOSM was designed specifically for OpenStreetMap data creation so it’s not an interface that city staff or surveyors would likely be familiar with, and the digitization process involves a couple of workarounds.
This post aims to break down the problem of curb inventories to illustrate what’s required, what issues cities should consider, and what types of options are available to them. While this isn’t an exhaustive list of tools, it aims to illustrate different pathways that cities can take, and to show that the problem itself isn’t prohibitively difficult — curb regulations are complicated, but they can be surveyed just like any other street asset.
If you’re interested in collecting curb regulation data for your city, we’d love to hear from you! We are happy to share our learnings so far and help if we’re able.