Geospatial storytelling, geo-evangelism, and how to build the ultimate reality capture kit.

By Pete Kelsey and Gavin Schrock, PLS

The geospatial sector has a bona fide “geo celebrity”, and that distinction is well deserved. Pete Kelsey has been out there in the geo trenches with the rest of us for decades. Beginning with military service, then an early adopter, expert, and trainer for CAD and civil software; Pete would go on to success with several companies of his own in the AEC sector. All along his journey, Pete evangelized the virtues of geospatial tech and best practices; eventually holding the title of worldwide technical evangelist for a major spatial software company. But his journey did not end there; what he’s into now is both fascinating and inspiring.

I remember, many years ago, running into Pete in the corridors of a CAD company; he stopped me and said with great excitement, “You gotta see what people are doing with laser scanners!” Not one to ever rest on his laurels, Pete began the next chapter in his storied life, on the vanguard of reality capture (RC); not just evangelizing but doing.

The geospatial sector needs more evangelizing: internally to keep practitioners moving forward and using best practices, to young people considering career choices, and especially to the general public, to let them know just how essential these solutions and services are. Unlike some in the sector who evangelize with generic “inspirational poster” insights, Pete comes from the perspective of first-hand practice. His focus on RC for storytelling has positioned him (no pun intended) as a go-to guy for spatial elements for features produced and/or aired by National Geographic, Discovery, and History Channels., and more. Pete is the owner of VCTO Labs, an RC firm with a spatial storytelling specialty.

We’ve had many blue-sky conversations over the years. For the most recent, I wanted to engage Pete in a thought exercise: what would be the optimal RC kit? Note that any products discussed are not endorsements, but examples of what there is out there, and there is a lot of great gear out there. 

GS: I remember the early days of laser scanners, when the technology was mostly being used by archeologists, for historic perseveration. There was a decade of “Hey, look what we scanned” type stories, which did inspire folks in surveying, infrastructure, and AEC sectors to begin using the tech. But it wasn’t until the more recent RC era that the ability to capture and combine the outputs from an ever-increasing array of sensors has matured to the point where they are not only useable but becoming essential. Sorry for the long preface; but what I’m trying to say is that the kinds of high-profile projects you’ve been doing demonstrate the full potential of RC. Be it for a historic preservation project, the rich 3D worlds you paint are also engineering quality and suitable for untold downstream uses.

In your presentations, you often say: ‘Name me a medium that provides context better than reality capture—you can’t’. It’s like the difference between trying to convey a full story with only the written word and adding graphics, illustrations, and photographs. Using spatial representations to tell stories, to present data, is now becoming imperative.

PK: Everybody knows data is king. Everybody knows that data without context is useless. So, the evolution of that is: name me a data type that provides more value. I don’t think you can, I’m dead serious, and I’ve started saying that at events, keynotes, and telling the audience, do you have any idea what we’re into right now? We’re talking about the most repurposable, valuable data type in existence.

GS: The recent xyHt cover story comes to mind; your full capture of Alcatraz.

That’s visually stunning of course, but more important, it is rich and precise data. I love the du jour developments like Gaussian Splats, NeRF, etc. Yes, realistic visuals can help make better-informed decisions, but the surveyor in me says: ‘What the *bleep* else can I do with it?’             

PK: People get that, you know, they understand, the power to do things with it is infinite. What can you model out of it? What attribution can you attach to it? How often can you update it? You can turn it into an actual digital twin. There are lots of ways to capture that; and with the tools out there, there’s no excuse not to do it right.

GS: Presently, the talk about AI seems to be taking all of the air out of the room, while other significant developments get little airplay, like RC. The AI hype cycle might be tempered a bit as some of the generative tool graphics start to look a bit dated, and news that, for instance, the limitations become more apparent. Did you see the Apple paper that says there’s a lot of amazing pattern recognition going on with AI, but not what would be considered ‘thinking’? ( bit.ly/4jQcF4o ). It would be great to see more talk and awareness around RC.

PK: I’ve caught myself saying a few times publicly that I love AI. It’s not going away. Whether you’re afraid of it or worship it, you know, whatever… It’s here. But I would argue that you should watch RC. Watch it move up and evolve rapidly. In the near term, there may be a more significant impact ahead, for certain uses, from RC, than certain types of AI tools.

GS: It’s like what I see about another buzz term “quantum”. Quantum computing gets all the airplay, but it is quantum sensing tech, like that which could be integrated into RC systems, that will have a greater immediate impact, despite flying under the radar ( bit.ly/4n3VejA).

PK: Exactly. Something RC has going for it is the visual impact; it really draws attention. Watching the Alcatraz project go completely berserk in terms of reach and people interested in it, I wondered, why are they interested in it? Well, obviously, it’s a great story. Everybody’s familiar with Alcatraz. Well, if that’s the gateway drug to what I’m talking about, to help promote RC for other applications, then all the better.

GS: Go forth and capture! But a question we often hear is ‘what with?’ Trying to keep up with the cornucopia of new and updated RC tools, sensors, systems, software, platforms, etc., is a blur right now. And add to that the new systems (popping up almost daily); many are simply integrating some commoditized OEM components. It’s tough to try to keep up.

That being said, here’s a thought exercise. Say there’s a mid-sized engineering-surveying firm that wants to build out a full RC team. I’d thought about saying ‘if money is no object’, but it always is, no matter what sized firm.

PK: I always start with the questions that any service provider in the RC space should ask. In no particular order. Let’s talk about the level of detail (LOD). Does it need to be tied to the ground? Does it need to be spatial, and measurable? Another one is, what’s this for? These pre-qualifying questions always determine the gear that you’re going to use, how long it’s going to take, and how much it will cost.

GS: Yes, depending on the desired or required LOD, fit-for-purpose tools and approaches can be considered. But for this exercise, let’s say it’s for high LOD applications, like civil or vertical design and construction.

PK: First, a word of caution. eBay and Craigslist are full of laser scanners and drones, etc., from people who have jumped into RC without making sure they will have enough work coming in. None of the RC kit is cheap. This goes back to land surveying; none of that gear was cheap. If that gear is sitting on the shelf, management is going to raise hell if it’s not out there in the field, billable.

GS: The business side of RC would make a great series of articles. But for argument’s sake, let’s say that someone has done due diligence, worked up a solid business case, and has the support of management to ramp up an RC shop.

PK: I’m going to give you my dream list. The projects that I’m doing involve everything:  interiors, exteriors, aerial, sometimes maritime, sometimes underground. So, in a way, I have had a bit of a unique series of experiences with all these different, sometimes really out-there project engagements. In the context of 3D capturing anything, here are suggested tools and methods. If it involves say, boundary surveying, that is a different thing, though RC tools can be used for certain aspects, like to help with the mapping aspects of ALTA (land title surveys).

GS: Let’s start with a summary list, then we can drill deeper into each.

PK: You’d want a SLAM unit and a terrestrial laser scanning unit. Then a full suite of airborne. You know, drone-based components and payloads. That could include infrared, multispectral, cameras, and lidar of course. For maritime and hydro work, a USV (unmanned surface vehicle) with multi-beam for inshore—I don’t know why anyone bothers with single-beam anymore—and things you can put on a boat for offshore. For underground, if the passages are large enough, you can use the same tools as indoor capture, or you might have to try ground penetrating radar, though that has a lot of limitations.

GS: Airborne, especially drone-based has more options than one can keep up with. But what do you consider when choosing sensors and systems?

PK: It is a whole big can of worms, and like any aspect of RC, there are key questions to ask upfront. LOD of course and how big is the area? What about beyond the visual line of sight? What about all the FAA regulations? Typically, the bigger the site is, the more negatively affected LOD, but that depends on tools and methods. Images, photogrammetry, spectral…?

There are systems out there where you can take your beautiful mirrorless SLR camera, mount it to a gimbal, and you can fly it— cool. What can make a lot of sense is sensors that can be multi-purpose. You’ve heard me talk about one that I use a lot, the Hovermap [Emesent]; I can walk with it, stick it on a drone, put it on a vehicle. I could put it on a boat or USV. The scanning tech is solid, a 32-channel Hesai scanner head. There are some other great systems out there that are also compact and light enough for drones, mobile, walking, boats, USVs, etc.

GS: Lidar on drones and other platforms can range from low-cost up to top-tier components, like Riegl scanner heads. Nice that we’ve got all of these options. And I’m glad to see more options now for the hydro/maritime applications.

PK: If you want to talk about what can be carried on a boat, the sky’s the limit. We had a recent project in Hawaii where we put a Leica TRK 700 on a boat that also could have had multi-beam, capturing the shore and underwater features at the same time. You could end up with say, $2M into the boat. That’s the high end, but quite impressive to show what can be done.

GS: I’m encouraged by what’s happening in the USV space. It seemed like, for decades, USVs were kind of frozen in time, the same funky-looking contraptions. But now we’re seeing some great new, sleek, and designed better for different navigation conditions. USVs with multi-beam, portals for additional sensors, built-in GNSS+INS positioning, different levels of autonomy, etc. Like the Apache, or Duck, nice to finally have more choices.

PK: We’ve used some of the highly autonomous systems for surface and underwater. The UAVs (underwater autonomous vehicles) kind of look like torpedoes; you set them on their way and hope they come back.

GS: Also, seeing some cool developments for shoreline mapping. Typically, in the past, you need a boat offshore, a USV for inshore, and a drone or something terrestrial for the land component. As you mentioned earlier, green lidar can now see down 20-plus meters depending on turbidity. There’s even green lidar that can fit on large-platform drones. But what really blows me away is hybrid, red, and green systems for aircraft mapping of all three environments at once.

PK: Underwater is getting better, but I wish we had more options for underground;  it’s still a huge challenge. GPR (ground penetrating radar) really isn’t my thing. I’m not a geophysicist. I have used them in the field with limited success. It can be useful; it’s spatial, it’s measurable, and I can integrate it with my point clouds and measurements. I’ve done that. It’s super cool, but what I can’t stand about it is that the data, to me, a parabola looks like every other parabola.

GS: Yes, GPR can only ‘see’ down a few feet, because of power limits set by agencies like the FCC in the U.S. (Federal Communications Commission), this is due to hazards of higher power emissions.

There are those legacy GPR parabolas that bounce back, and making sense of them is like reading tea leaves. But things are changing. While a lot of the lower priced lawn-mower style GPR systems have only one or two sensors, there are larger systems now with 30 plus sensors. These are push-carts or vehicle towed. And with those arrays of sensors, it can detect in, sort of, 3D. There’s even some software that does a decent job of mapping pipe diameters and courses. You should check out the new wave (no pun intended) of GPR systems; hopefully, the software can help demystify those images of signal returns.

Speaking of underground. What are some cool options for indoor and larger underground spaces you’ve used?

PK: The Hovermap ticks nearly all of my SLAM boxes. For indoors I can carry a SLAM unit like that, put it on a cart, and if the indoor space is large enough, even fly it on my drone. But, sometimes the indoor passages are a bit small, or there are pipes or ductwork. What I really like about that is something like Elios 3 from Flyability, a drone in a cage. Alcatraz is a great example. We couldn’t put our drone in the sewer lines, smokestacks, and cisterns. The Elios 3 has a 4K camera, a light source,  and LIDAR. And because it’s in a cage, it won’t crash into anything; I can get it back.

GS: Drone solutions for indoors, especially for mines, is where I’ve seen the first of the level-4-autonomy drones, like from Exyn ( bit.ly/3G0buBD ). Most drones are level 3: you define, or have the software define flight patterns or set waypoints. When you get to level 4, you only have to set some volumetric limits, and the drone just keeps exploring the passages until it runs out of places to map.

For outdoor drone operations, what are some considerations for lidar? I mean, I’ve been impressed with even middle-tier lidar and photogrammetry demo data from wide-open places, but here in the Pacific Northwest, we have some mighty dense foliage.

PK: What I’ve learned is, if I want to be able to scan anything. I want a five-return laser. It might be a bigger investment up-front, but then you’ve got more of a chance of success in more places. There are lots of drone lidar outfits such as Phoenix Lidar that integrate multiple return lidar heads, 2, 3, 4, or 5 returns, like Riegl makes. With a Riegl Phoenix LiDAR five-return laser I can punch through triple layer canopy in the Yucatan Peninsula if I need to.

GS: You know that I tend to get fractious about positioning precision. I’m a surveyor; I position, therefore I am. It is encouraging to see manufacturers in the drone, mobile mapping, and even SLAM space getting much more serious about their positioning stacks. In the early days of mass-market drones, they used cheap GNSS, I’m happy that there are developers integrating top-shelf GNSS boards, IMUs, and software. Inertial Labs has always impressed me in this regard, with good NovAtel boards and the Waypoint Inertial Explorer for PPK.

PK: Through my experiences I’m at the point where if the client says, ‘No, I don’t care about position’, I’ll do it anyway. That tells me they’re not thinking clearly. I know, in my experience, that it will work out better for me if I do it anyway because the alternative could be, six months from now, they call me up, screaming, yelling, ‘How come the data isn’t spatial?’  And it’s not an expensive component to do right.

At the bottom of this pyramid of RC, is all this cool gear and software. But the foundation of that should be the most bad*** GNSS gear you can afford. To keep your spatial component as tight as you can.

GS: Definitely. I do though see folks go super cheap on the GNSS gear, and that can backfire. You need a good rover to set ground control points, and to collect static for PPK post-processing, or to broadcast RTK corrections. However, if folks really want to get the best results, a mix of GCP and PPK is the best bet, and not just relying on RTK alone. Sorry… don’t get me started…

I guess the same applies to GNSS gear as with other RC sensors; under optimal conditions, super-cheap gear can work. You might not have to always go for the top-tier gear, but there are a lot of good options in the middle ranges.

While we are still discussing drones, we touched on platform size, but what about platform type?

PK: There are so many rotor drones that can handle most small payloads, like the various DJIs, but great domestically produced drones like Inspired Flight, Freefly Astro, etc. Then you have larger platforms, like the Freefly Alta-X and others. You have the flexibility of carrying more types, sizes, and weights of payloads, but you have to have some skills; they’re pretty big and could seriously hurt someone if they crash.

GS: Well, there are the parachute systems. But that aside, makers of large platform drones have usually sunk a lot of R&D into failure prevention. For example, Freefly built drones (many years before they got popular), for the film industry, where you would not want to crash while carrying a quarter-million-dollar camera.

PK: Pretty much, a large platform drone ensures that when a customer comes up and asks if you can do something, you can. But not in every situation. Sure, rotor drones are the most common, but if you want to offer services for long corridors, especially beyond visual line of sight (BVLOS), a VTOL drone may be the only practical option. Unless you want to segment small courses for your rotor drone–which is not very efficient.

If the customer says they’ve got 40 miles of pipeline or power transmission lines, or thousands of acres to map, and they don’t want to go to the expense of piloted aircraft or need something higher definition from a drone, I’m impressed by VTOL drones like the Centaro 5 or 6 from Censys Technologies, out of Florida.

GS: What about terrestrial laser scanners (TLS)? I mean that SLAM is super useful for many applications but can’t meet every LOD need or requirement.

PK: Sure, for indoor especially, if you need high LOD, units like the RTC360 are great. Workflows for those, and others have really moved on past the old days when we had to set all of those ball targets. These have amazing, automated registration options.

One thing I’m keeping my eye on is the new Riegl V600i. It’s as compact as a lot of other TLS, but with a range and output that is a bit surprising for the price. And it’s a Riegl! They’re sticking them on cars, carts, and backpacks; it will be interesting to see where this goes. 

Indoors, modeling for design, construction, inspection, architecture, etc., really needs high LOD Outdoors, that might not be the case as often, at least not to where you couldn’t do most things with a drone or handheld, or mobile.

GS: I do see large format TLS being used less, although there are applications where long range is needed. There are other sensors that can, for short-range applications of less than 30m, do a lot of what folks used to rely on the big ones for.  I know that this raises a lot of skepticism, but I recently tested a handheld, multi-camera-based handheld, the output of which had a denser point cloud than a huge scanner. We compared the point clouds with a TS scan of the same sites and could not see any difference.

There’s a lot going on in how tightly integrated sensors, with advanced algorithms, including some AI, are rewriting certain past assumptions. This is a good segue into another class of system: mobile mapping.

PK: Hands down, right now, the Leica TRK 700. There are a lot of systems with great scanners, good positioning, options for good cameras, and lidar, all the way up to Riegl scan heads. I’ve used it on projects, and it captures almost everything we need along a road, or in the case of the Hawaii project, the shoreline from the boat. Great point cloud and images.

GS: Well, I am not saying this is an endorsement but when that first came out I was a bit awestruck. I dove into the details, looked at a lot of data, and even did some test drives. There are great systems out there, but at the time the TRK really stood out. From the innovation in approaches to the positioning stack, for instance, SLAM position stabilization, to the scanner and camera options, and in-system image anonymization.

However, mobile mapping is profoundly data and processing-intensive. For example, the TRK has a processing box you put in the back seat, and it begins the processing even while you are still driving, a lot of larger mobile mapping systems pack processing boxes. On that note, what about processing approaches for your ultimate RC kit?

PK: Well, to your point, one of the first things that comes to my mind is the notion that some folks preach, that the cloud is supposed to save us all—no, it won’t!

Case and point, this is surveying, land surveying 101. You pack up everything, you go to a remote site, you do your capture, your survey, your measurements, whatever. And everybody’s nightmare is when you get back to the office and realize you didn’t get it all. Or something’s corrupt, there was a technical problem, and you have to go back. That’s a reality, and probably always will be.

Even if the cloud seems to be available at a particular site, it likely won’t be at others. Alcatraz was a perfect example, the Wi-Fi seemed like it was from the 1980s.

GS: I don’t think we had Wi-Fi in the 1980s…

PK: Exactly. We need to pack processing capabilities with us, to make sure everything is complete before we leave the site. Sure, we can sync up with the cloud later, but this type of edge computing is essential. I’m doing an event with AMD soon, they are really embracing the RC thing, they want to be the brains behind RC edge computing systems.

GS: When it comes to engineering grade RC, surveying for engineering design, monitoring, and construction, RC sensors may not be able to capture everything. And there will still need to be some discrete shots taken with a total station or GNSS rover, like utility inverts, control points, etc. It is interesting to see how popular scanning total stations have become, more in Europe than in the U.S. though. They are usually a very good total station, that can also do selective, albeit slow, scans and imaging over short ranges.

But that aside, if you were to pick one essential piece of RC kit, the most versatile for a broad range of potential projects, what would you recommend?

PK: That would be a SLAM unit, But a good one. Like with a 32-channel scanner, a good positioning capability, maybe with a camera or cameras. You can stick it on a drone, a vehicle, a boat, or a backpack. Yes, there are limits for LOD, but some are really getting better at that.

GS: Then to a point of pain we hear from folks that the era of ‘scan everything!’ is creating bottlenecks in their operations. An RC R&D engineer recently told me; the reason was obvious:  software hasn’t caught up yet. Until we can further automate what we need to do with the data, there are ‘office’ backlogs.

PK: I couldn’t agree with that more; it’s an elephant in the room.  Oldsters like you and I, know that data sets never have and never will get smaller. They’re only going to get bigger and bigger. True with DWG, DGN, geodatabases, point clouds, everything.

GS: So, we hope and pray that the R&D folks can create tools in the software to streamline our office loads. Meanwhile, the tsunami of RC data continues to grow. More data is valuable if we can handle it, but should we also be reevaluating the downstream segment as well.

PK: Can we cut the cord with CAD? Now, let me tell you what I’m talking about. Everybody in the capture space knows because we’ve been doing this since it began with the very first point cloud. Well, can I bring it into CAD? Can I bring it into BIM? Yeah, but you’re going to have to decimate it. You’re going to have to dumb it down. You’re going to have to blah, blah. blah…

And to me, it’s just like, Can we move on? CAD is the same as it was 40 years ago. Lines, arcs, circles, text. It hasn’t changed, and it just feels like some super bright young startup folks out there could come up with a new way to do the downstream work without dumbing down point clouds. Because all we’re doing is losing fidelity. Can we? Can we cut the CAD cord and create something else in the captured data?

GS: I agree, but small steps. The built environment has defined surfaces and features, and for design and construction, there would need to be some modeling. However, to your point about preserving the true 3D environment, there are some pretty exciting things going on. For example, when Bentley announced the acquisition of Cesium, this opened up so many possibilities to preserve the captured data in a digital twin environment.

We could go on for hours on this subject; perhaps another day [Note that we had already been asked to leave the Sushi bar we were in, for sitting there talking too long]. Any last tips or warnings about common pitfalls?

PK: Cost. You know, be careful. Reality capture presents very large startup costs. So make sure you have the work. Make sure you get trained. You know, any 18-year-old can spend $1,000 on a drone and go create photogrammetric meshes, etc.

The human component, the team, you’ve got all this great kit, but you got to have a team. One or more are going to have their 107 (Federal Aviation Administration drone certification), but the drone and other gear have to be skillfully operated, maintained, and calibrated. You might have to have a licensed surveyor as well, depending on the type of work. You have to deal with permits. If you do mobile mapping projects and drone projects, you need someone who can navigate the permitting process.

It’s not just like you can just hire a bright kid, buy a drone, and press ‘go’. It’s not safe and your gear might end up on eBay or Craigslist.