I Saw the Sign: 3 Trends in Traffic Sign Inventory with Street-level Photos
Keeping an overview of the traffic signs in an area is not an easy task. We’ve learnt a lot from our customers in this field and want to share 3 trends in harnessing technology for traffic sign inventories and asset inventories in general: 360 views, cameras offering competition to LIDAR, and the “Daily Street View”.
If one spent a minute recording the location of each traffic sign in England on a map, it would take over 25 years to complete the inventory. Even at the scale of a small city or town, our customers tell us that traffic sign inventory is "months of painstaking work collecting street sign locations manually". As we’ve been working with people in GIS, Public Works, and Transportation over the past year, here’s what we’ve learned about the workflow.
- It costs $4 - $6 per sign to have inspectors physically go on location to inventory signs.
- Many try their own flavor of “DIY street view” solutions when Google’s coverage isn’t recent or complete enough.
We see three clear trends coming out these efforts.
1. 360 is the new smartphone
These days it’s going to take a lot more than a mobile app to impress a GIS Analyst. Most of our customers have government-issued devices or tablets. Many of them are already using ArcGIS Collector or Fulcrum as a means of snapping photos in the field for reviewing later.
But sometimes one vantage point isn’t enough. 2016 saw a flurry of new consumer grade 360° cameras come out on the market, which then turned into a flurry of questions that came our way. What are the 360° cameras on the market? Can the panoramas be stitched together like Street View? Can you fake it with multiple GoPros?
Here’s what we know: There is a long (and growing) list of affordable 360° cameras. If you’re looking to use one for traffic sign inventory, the best cameras for this have decent image quality balanced with the capability to take images at a relatively fast rate (ideally less than a second) and don’t require lots of post-capture processing.
Computer vision technology makes it possible to create your own “Street View” from the series of photos or panoramas captured. A technique called Structure from Motion can stitch the photos into a navigable experience based on matching features within the image. This also works for images taken by multiple action camera setups for simulating a 360-ish experience.
Multi-cam shots of San Francisco, CA
2. Camera’s rise alongside LIDAR
For customers who can afford it, LIDAR is still the winner for high-precision, asset grade surveys. Most mobile mapping companies use professional-grade 360° cameras to complement the LIDAR capture, for locating traffic signs which can then be placed in the LIDAR 3D point cloud with centimeter-level accuracy.
For customers who can’t afford and/or don’t need that level of precision: LIDAR is overkill. With good GPS and plenty of photo overlap, Structure from Motion can do a “good enough” job of creating 3D reconstructions from street-level images for estimating locations for traffic sign inventory.
Our partner Continental Mapping provides high-accuracy mobile mapping surveys with high-resolution 360° images
3. “Daily Street View”
Given the lower cost of capturing imagery with “good enough” results, multiple agencies from different parts of the world have independently approached us with the same Big Idea: is it possible to gear up municipal fleets—e.g. sanitation trucks, field work vehicles, parking enforcement cars—and collect street-level imagery every day?
Equipping municipal fleets with cameras to capture a "Daily Street View"
While the breadth of data extraction possibilities from “Daily Street View” is truly daunting, it begs the immediate question of whether it’s really useful for traffic signs, which don’t change frequently. We didn’t have to look far for the answer(s).
What we’ve heard from the field is that having a well-maintained database of traffic signs makes for better protection against claims (a sizable budget allocation for U.S. agencies relative to their counterparts in Europe). It’s also a good practice for disaster recovery; both for providing a recent enough pre-event visual record for property damage claims and for assessing the level of damage directly after a storm or fire.
The common thread across these trends is smart use of your own photos as an alternative to Street View, made possible with recent advances in hardware and computer vision. As we expand our detection capabilities to more objects, we’re excited to see our customers lead the way to better solutions for asset inventory beyond traffic signs.