Welcome to our new post series featuring RoadBotics’ Research and Development.
First up, we’re talking about 3D reconstruction – what it is, how it works, and how our team is making strides to incorporate it into our assessment technology.
The term “3D reconstruction” covers a lot of ground, but this post is meant to be introductory in nature.
What is 3D Reconstruction?
Everyone in the modern era has exposure to computer-generated 3D worlds, and a massive amount of effort in the hardware and software industries goes into making the graphics strikingly photorealistic. For example, the latest demo by NVIDIA to showcase their latest video cards is stunning:
But this is an example of a constructed world; it doesn’t correspond to a real room somewhere in NVIDIA’s headquarters or anywhere else.
3D reconstruction, broadly speaking, refers to techniques and technologies meant to provide a digital 3D representation of something in the real world. (Thus the “re” in reconstruction).
Digitizing museum pieces, such as sculptures, was an early practical use case for the technology and remains one of the most accessible uses. The Smithsonian Institute was an early adopter of the technique and used it to create some very compelling 3D models that are reconstructions of real objects in their collection, like this ewer:
When successful, this technique has a myriad of benefits beyond appreciation and edification. One of the obvious benefits is that software and technologies already developed for 3D constructed scenes and objects can now be applied to objects that are captured from the real world. For instance, a common commercial use-case is to capture a person’s body and facial features, and then use that model in a video game. In an engineering context, one might take a 3D reconstruction of an object and further manipulate it or test it using CAD software.
Why RoadBotics is Pursuing 3D Reconstruction
Simply put – we think that efficiently producible accurate 3D models of civil infrastructure will be the next great GovTech revolution. 3D reconstructions will make data-driven decisions easier and cheaper than ever before, allowing towns to make increasingly sophisticated judgments about their infrastructure without needing very labor-intensive and expensive CAD-type models. Just as RoadWay Pavement Assessments are a giant step ahead of old pen and paper sampling-based assessment methodologies, capturing 3D models of infrastructure cost-effectively and at scale will allow us to do things that governments simply can’t do right now.
One use case we are particularly interested in is automated sidewalk reconstruction.
As many of our blog readers know, the Americans with Disabilities Act (ADA) mandates that sidewalks must meet certain design criteria regarding their safety and quality, with the goal of ensuring accessibility to all Americans. Ever since the ADA came into effect, local officials have been hard pressed to both retrofit existing infrastructure (e.g. by building ramps or widening pathways) and maintain it (e.g. by removing overgrown vegetation or repairing structural damage that can pose a tripping hazard).
These requirements were created to increase city accessibility to people with disabilities. Municipalities of course want to allow their constituents to have free mobility and access, and routinely spend large amounts of money and effort trying to comply with the ADA.
Before such improvements can be initiated efficiently however, city managers and public works directors need to know where investments are needed and how much work is required.
The baseline requirement is, of course, data. Specifically, a complete and up-to-date inventory of a community’s sidewalks is a crucial tool when making these kinds of decisions.
Traditionally, such inventories require costly manual inspections and can take many months to complete, or rely on pedestrian feedback, which is prone to errors and is unlikely to deliver the comprehensive data needed to make an informed plan of action.
Even sidewalk inventories, as crucial as they are, have limitations; they can only give visual information about sidewalks. This means you can’t do more detailed analyses that you could do with a 3D model, like measuring parts of the sidewalk.
For instance, take sidewalk width requirements.
The ADA generally requires sidewalks to be at least 36” (3 feet) wide at all points. In addition, if a sidewalk is not 60” wide (5 feet), then it must have passing sections every 200’ that are 60” wide. It is time-consuming to manually ascertain whether a section of sidewalk is compliant with this standard, much less all sidewalks in a whole city (I.e. by having an engineer literally measure the sidewalk.)
The width requirement of course is not the sole requirement of the ADA – there are also numerous other requirements such as the slope of a sidewalk, curb measurements, and more.
With an accurately reconstructed 3D model of a sidewalk, one could digitally measure the sidewalk’s real-world width in seconds to find out if it was ADA width-compliant. This would be overkill for a single measurement using traditional reconstruction methods, but we are working on making it just as easy as self-collection using RoadSense. And these measurements (and others) could also be derived automatically, at scale, and in a fraction of the time it would take to manually perform the same inspection.
Imagine having a RoadWay-style map made up of actual 3D models of every sidewalk in your community, and that you could measure, export, model, and more!
Now you see why we at RoadBotics are so excited about our 3D reconstruction technology!
In the next installments of this series, we’re going to delve deep into the state of 3D reconstruction technology, the problems we’re working on to bring our 3D reconstructions to a product for you, and a sneak peek at our accomplishments to date – stay tuned!
