More than Just Fun with Drones: Producing Engineered Products

ERIC CYLWIK: Well, good afternoon. My name is Eric Cylwik, and I am with Sundt Construction. And I'm a virtual construction engineer. And despite what I look like, I've actually been with Sundt for a little bit over 11 years already.

And my background at first, during high school, was in IT. And then I got my degree in design studies with an emphasis in digital visualization, and started out at Sundt just creating animations to show means and methods for the construction process during CMAR, CMGC construction interviews to be able to really communicate why Sundt was the best value provider for a particular contract.

And then as time went on, I'd go out into the field and observe projects that we were selected for. And I'd always find people having these discussions about, hey, how do we do this? And I'd be like, well, I have a 3D model of it. And they're like, what? I'm like, well, yeah. That's what we told the owner we were going to do, is this great idea. And you guys are figuring out how to do it again.

And they're like, can we see the model? And I was like, yeah, for sure. And then I was like, yeah, and it's at scale, so I can give you dimensions too. And they were like, what?

And so I continued to transition more over to the operations side. And so now I focus probably about 60% of my time on the operations of our construction job sites and support that from all aspects on a virtual design and construction and also operational technology side. So I get to play with a lot of fun toys, but at the end of the day, I get to say, hey, I'm proud I built that bridge or that research institution.

So today we'll be talking about drones. Sundt has had an interesting journey so far with drones, and we've been pretty lucky to have already had some experiences on the non-drone side of what we wanted to do. So that way when we saw this technology arise, it became a pretty clear path forward of how we could add value to our projects to make sure that we had actionable data.

We didn't just want cute little photos of our buildings. We were already getting those from other places. So how do we equip our teams with another layer of technology that makes their job better and faster and easier? And there's just so much communication and information that needs to be shared on a construction site.

So today we'll start off with a little bit of an introduction about how Sundt got to where we are today with our drone program. I feel like we've grown pretty drastically. I know in comparison to other companies, we haven't been quite as drastic. But I want to cover that. There's a lot of good information there.

I want to talk really briefly about what makes data engineered. And it's a custom definition that made just for this course, so it's perfect for you guys. And then we'll talk a really quick crash course on photogrammetry. How many people have not heard of photogrammetry?

Awesome. I will spend just a moment or two on it, and that's what I was hoping for. If you want to pull up your phone and Google it real quickly, I won't judge you. How about that?

And then we're going to have a use case. So actually looking at, hey, we've taken this engineered product now, and we're using it as part of our construction deliverables and using that to communicate amongst the team. And then I wanted to touch on two subjects-- two pieces of technology that I would consider future state of technologies that are just now emerging.

And not many people have adopted yet, but I think they are really going to make this idea and this concept much more streamlined and much more widespread in all of the industries in construction, and also all the adjacent industries. And then we'll wrap up.

So talking about the evolution of the unmanned aerial systems program at Sundt, it started around 2015. We had two pilots and two drones. One of them was a hobbyist that just loves buying new things and pushing the limits. And he started taking photos of some of our job sites. And then, in that same period, we also hired somebody who had been doing some third party missions for just random people around town, taking photos and getting into it to be able to show progress.

And so at this stage, it was, hey, this is a fun technology. Let's take some photos and see what we can do. And it started out with two Phantom 3s. The following year, we ended up having five pilots and three drones. We purchased a Phantom 4 Pro. And then we began to experiment with photogrammetry, being able to take these things and start to produce something that was usable in other platforms other than just viewing a picture on a smartphone or other device.

In 2017, we had seven pilots and five drones. The one up here is the GoPro Karma, which is our only non-DJI model at this point in time. And we began to consistently provide 3D models. So you can start to see how this is scaling out. Hey, this one pilot liked it. He got somebody else trained on it, so that way there was some backup on a job site, which is why you'll see more pilots than drones.

And then also, you'll see, hey, a little bit of experimentation and then people starting to trend towards that. And we're also providing to-scale imagery. And I'll dive into that just a little bit later if you're unfamiliar with that term.

But then in 2018, we made a first heavy investment into our drone program in the form of a Matrice 210 with digital RTK. It's a semi-industrial class drone. It's capable of flying at power plants and around high voltage lines, where GPS loses signal and drones go a little bit crazy. Thing's able to keep its game face on.

One thing that's nice about it is that it's got two camera mounts on the bottom. So we're able to take both a color, a normal photo, as well as a thermal image. And with this drone, we've been able to provide extremely high precision models, images, as well as start to dive into the realm of looking at what we're doing with envelopes on a building from a thermal perspective.

When we turn a project over to an owner, our goal isn't that we get called back in two months because we didn't seal a window properly. We want to know that before we turn it over. So we're proactively looking at what we're turning over and understanding the performance of that structure before we step foot away from that site.

One of the other nice features that came out is DJI now does on the Phantom 4 Pros, 360, or spherical, photos, which are fantastic for documentation. It's a photo-- if you can imagine being able to pan around and look anywhere you want. A lot of times when I'm showing people photos of their job site, they think I've somehow mounted a permanent drone live webcam. And I tell them, yes, I'm watching you all the time.

But so again, we started providing actual engineered deliverables in 2018. And I'll, again, define that in just a moment. But then also, starting to do thermal mapping on slabs immediately after a concrete placement, and then also looking at envelopes for building performance.

And then also, we jumped onto the Skyward platform. And if you're unfamiliar with them, they're a very forward-thinking company that's into fleet management and safety management. So as a construction company, we don't just go out and say, hey, I know you normally do wedding photography with your drone, but why don't you fly near our tower crane? Bad idea.

So Skyward is a platform where third parties can go. They get to log what missions they do, what kind of activities they perform. And it allows us to have a bar that we can set to say, hey, if you've never flown a construction job site, I don't even really want to know that you're a potential provider for this service.

And then it also automatically logs all of our pilots' flights. That way we can make sure that the FAA regulations are being followed. And looking forward, that's an extremely important leg of our program to make sure that we have organizational consistency, while also equipping the front lines of our business with the technology that they need to be able to better serve our clients.

So as far as engineer deliverables go, this is actually one of our first projects we did. We had a 250 acre water retention site, and the owner wanted us to do an as-built deliverable every three months to be able to justify earthworks billing. And the preconstruction topo, two people went out and did it, and it took about 90 hours for each of them-- 90 hours plus GPS, and truck, and all that stuff.

And then the second time we did it, we had a third party provide a drone service. And they went and flew the site in four hours, and then they spent two days processing that data. So three labor days and then four hours of drone time, and the amount of accuracy we got was unbelievable at that point in time.

I remember turning the deliverable into our survey manager. And he opens it up, and he's like, yeah, yeah, this is garbage. I figured it would be. They don't have this drone stuff figured out. And I was like, oh, what's up with it? He's like, well, the contours-- they're all like jagged and messed up. You can tell it's trying to guess at stuff that isn't there.

And I was like, oh, let's zoom in there and take a look at the high resolution data. So I opened up the point cloud and zoomed in. And what he was looking at was-- he was expecting a nice, smooth contour along the slope, which is what you would normally get if you did a 50-foot grid with the topo. But he was actually looking at the scraper tracks that were running that day, and deciding that was bad data, because we had that level of detail.

And that's crazy that not only did we increase the quality of the deliverable, but we also shaved a whole bunch of time off of it. And from my perspective, we also shaved a whole bunch of risk off of it, because now we don't have two people running out there with a GPS in between scraper hauls. That's a pretty dangerous spot to put a human. So hey, is there a better way to do this?

So that was our first exposure to it and definitely a learning curve for us. After going through that, we started to say, hey, how else can we use this idea of to-scale imagery? So one of the outputs from this is it produces-- it's called an ortho TIFF, is the technical term. Orthomosaic image, effectively.

And this is a-- if you can imagine there are probably 15 or 20 photos that were taken from this. And it stitches them together like a quilt, and it creates what looks like a satellite image looking straight down at something. So you don't see angles on walls. You just see the top of the wall.

But when we're doing a concrete placement on a post tension deck, these post-tension cables-- the red lines in this image-- are extremely important. For one reason or another, somebody always has to go back into the deck and drill or cut something to be able to play something after the fact. Oh, I forgot to put this sleeve in there. I need to drill.

Well, if you hit one of those post tension cables, they're under extreme stress, and they will snap and shoot out of your building. And that has a huge safety risk, because if there's somebody in the street, if there's somebody doing some work in a scissor lift on the side of the building, that's extreme danger for a steel cable to snap-- that's under compression-- and just rocket out the side.

So with this, if somebody wants to drill, we can now go in and measure on this image-- this is to-scale-- from the edge of the slab to wherever that post tension cable was. Because again, this is a straight down image. So we started to be able to access not just to a picture of something, but we could go in and give a real world relative dimension.

And that was a big game changer for us to be able to say, hey, historically, you might take a few photos of this. And then you'd go back and say, well, maybe here's a good spot to cut. And you go for it. Or you get a GPR system out there, and you scan the deck to try and identify it. But that wasn't always reliable, whereas this is a color image. You can clearly identify where those cables are.

And then the next natural step for us was to say, hey, we can do overlays. So in AutoCAD in this case, we were able to do an overlay of a PDF on top, and align it with that image, and be able to start to look at things and have discussions with cabinet installers about, hey, room 702 has a whole bunch of post tension cables running right under where your cabinets going. So while you're installing that, the concrete will still be wavy and dealing with the stress that's going through there. So you're going to have to make sure that you bring some extra stuff, maybe some extra shims.

But to be able to talk about the quality impacts of where these things are, because you're not really going to see a construction drawing that shows your cabinet overlay on top of your post tension cable overlay. Right? Those two things normally don't meet. But in this environment, it's extremely easy and intuitive to do so.

The other thing we started using it for was to be able to identify where sleeves or block outs are, or are not. Again, we can measure to these, understand whether or not they're within a specific tolerance. And then we can also identify, in this case, that the subcontractor installed some additional things, which doesn't sound like the end of the world. Maybe they're proactively hedging in an issue that they think might happen.

But on a project that's fully coordinated in 3D and all prefabbed, all of a sudden, you're like, well, what else is in that area? So then we can open up the 3D model, go to where this is, and then take a look and see if their new addition to what's in the real world affects anything that's in the planned or digital world.

And so we started being able to make actionable steps. And as we started to do that, those were all relative things. Hey, from the edge of slab to here is this dimension. Right? And we had a project that required engineered data. We needed to be able to fly a drone and get real world accurate measurements for the surface of the Earth, so a 3D topo.

And we needed a process that we could say was tested. We needed it to be predictable to say, we're going to fly. It's going to take this long. The accuracy is this, and I can prove that. And then it needed to be repeatable. And that doesn't sound out of the realm of ordinary for most survey activities, but for drone work, this is all new stuff-- a lot of it is. And it's really difficult to partner with a subcontractor that understands those requirements and is able to consistently and reliably provide that particular data.

So the crash course on photogrammetry-- this is from a screenshot from a program called Pix4D. Each one of these green spheres represents a photo that was taken from a drone. And then it projects all of those images to the point where they overlap. And it looks for visual artifacts-- so where there's a break in the curb or a crack in the asphalt. And it ties those things together and triangulates it.

And then you can click on any pixel here, and it'll show you how it triangulated that particular coordinate. So it's using many photos, triangulating a specific point, and then placing that point in space. And so the product is very similar to a high definition or regular laser scan, rather, let's say-- a point cloud type thing. And then you can go in and apply a mesh to it and kind of create some 3D models and stuff.

But ultimately, at this point in time, we hadn't been providing engineered deliverables. That was still a little bit out of our reach. So we partnered with a company out of Tucson named Phantom Aerial Solutions. And this individual got into the drone game a few years before we did, and previous to that, he was a heavy civil equipment operator, who started out as a laborer, worked his way, purchased his own equipment.

And then he realized how much survey was integrated with being able to do the right things on equipment. So then be bought his own survey gear, got a survey license, and was doing a one-man show. And then he was like, hey, I could just fly drones, and then I don't have to sit in the cab of a blade all day. So he went out and started this company. But he's got the knowledge, and the skill set, and the understanding of what happens on the construction site, and how drones can influence that. So it was an extremely strategic partnership.

The ultimate thing we're interested in is this project. About 30% of the budget for construction was to deal with a landfill that was under the proposed building. So we had to go in and dig down about 25 or 30 feet, pull out a whole bunch of garbage, go swap that with another landfill on most occasions. But if we weren't exporting garbage, then we had to go get it from a different location.

And so there was this crazy mismatch of all these quantities being exported and imported. And the owner, looking at this, was like, I've never spent 30% of my construction budget on site work. This is insane. How do I feel comfortable with this? And so we said, hey, we've got this technology. We can create this deliverable that we'll turn in.

And it's going to be a proven process. It'd be like going out there and doing a superfine topo every other week on your project. And ultimately, it helped the client understand what was happening, how things were evolving on site. And we were able to turn in bi-weekly quantity reports.

The guy in the back is going to start to yell at you guys in the back of the room for standing in front the door. So there are some seats kind of up here if you guys want to journey up a little bit. Otherwise, turn around and give him a threatening look. That'll let him know how you feel about that.

So here's-- I call it a cartoon of what we were trying to do. So here is one of the 3D models that we created from the flights before we started construction. This yellow area is the garbage that we had to export. The blue is a surcharge that the original contractor that was doing the landfill put on top of the garbage to cap it off. The red is just existing natural dirt that was there.

So those three quantity types needed to be tracked as we exported them. And then we had to bring back in-- for every yard of garbage we exported, we took it over to another landfill, and they gave us a yard of regular dirt from their landfill. So we needed to track that, how we moved the surcharge around on site, and then what we did with the natural grade.

So there were a lot of moving parts and pieces to this, and to be able to track all these things with just GPS data would've been pretty difficult. So we came up with this idea of, hey, let's go capture it with a drone. We get that aerial image. We also get surface and 3D data. We would then run that through Pix4D, in this case, to make all of those things come together and produce that engineered product. Pix4D is not the only solution out there for this. It's just the one that we happened to be using for this project.

The deliverable would then be pulled into Civil 3D, where we could look at the contours. We could look at elevation and slope data. We could compare it to previous flights to understand quantities. And then from there, it would generate out into a report that was based on Excel and some of the visuals from Civil. And then we also ran it through InfraWorks to be able to create a 3D mesh that's photo mapped and able to pull into Navisworks.

So we're going to take a look at this whole process here. So for capturing it, the bird that was selected was an Inspire 2. Some of the advantages of this are that it has dual GPS units. It also has dual IMUs, or inertial measurement units. And that allows each photo to be tagged with a higher precision of accuracy than just a normal drone that's a little bit more consumer grade.

Beyond just that, to rely on that for its accuracy, there were also four permanent survey markers, that were installed previous to construction, that were leveraged throughout the entire process. And those were made sure to not be disturbed, and they were visible in every single flight. And then for each flight, there were anywhere between 12 and 20 ground control points that were placed on the day of.

So some of the rules of thumb that the Phantom Aerial Solutions had are anytime that there's an elevation difference of more than 20 feet, that required another GCP, so that way the photos could properly be calibrated. And then he also just sprayed them out all over the site based on an average distance that he was able to identify. And those are extremely important things, because it really helps fine tune all those calculations for all the photos.

You want to try and pick spots where the GCP's on the inside-- so GCP is ground control point-- but where they're visible from multiple photos. So as you're looking in-- I'll show you the mission planning software here in just a moment. But you'll want to pick spots where the GCP is visible from multiple photos. That's how it calibrates the entire mesh. And then also, these ones made sure that-- the ones around the perimeter that were fixed and there on every single flight-- made sure that the mesh stayed absolutely true. Does that make sense?

So it's position in the real world never moved. Those yellow ones locked that in. And then all the changing site conditions as we were digging down 30 people or so were then held to their own by the fact that we had a whole bunch of ground control points inside of there. And those were all on a per-flight basis.

The flight control software that was used is a program called Map Pilot for DJI. And we did a whole bunch of research.

Yeah, question?

AUDIENCE: Real quick. On the GCPs, did you use all of those as GCPs, or were they used as checkpoints or 2D points--

ERIC CYLWIK: So we did three--

AUDIENCE: [INAUDIBLE]

ERIC CYLWIK: We did three flights before we started construction. The first one had just a whole bunch of GCPs all over the place. The second one had some GCPs as well as some checks. And then the third one had just GCPs. And we found that the checkpoints-- it's way more accurate to process this data, and two days later, go back out with the GPS and check it if we're worried about it. We're going to get a much more accurate readout from walking around on the surface, rather than just having a specific point identified in an image. Great question.

So in the software that you use to process this, you have the ability to take a mark that's an identified point, and it doesn't change the calculation. So it doesn't help calibrate it, but it'll automatically compare the difference between where it calculated that point and where you told it it was. So sometimes that's a handy check if you're willing to put in a little bit of extra time for that.

So this is a program called a Map Pilot for DJI. And one thing that's really great about this is it has a focus instead of just doing a grid pattern. This is a software that helps you identify mission criteria, and then it automatically updates your flight in real time based on what's happening with the conditions. So as the lighting conditions change, the program, in real time, identifies something called ground smear.

And that's the calculation of how fast the drone is moving, how big each pixel is on the ground. And then if your drone has to expose a single frame longer than that pixel is distance-wise, it will automatically slow your drone down. Because what will happen is you'll set up your drone, and it's running fine. And then a cloud rolls in. Well, those images either won't be bright enough for your drone to be able to calculate the same consistency for that, or your drone will allow it to expose that photograph for a longer period.

And then you end up with more smear. And now the computer's on the back end trying to identify, well, it's a circle here, but it's an oval in this one. And that really plays havoc on a lot of the way that these things identify and create these 3D models. So this software has some pretty advanced features to be able to go in and manage all of those things in real time.

So to show planning a mission, this is just a screen capture of an iPhone to plan the mission. So you go in. It's got a Google Earth style interface. And you start to press and hold to identify what area you want to fly. And then in this case, I'm doing a simulation, so I just tell it where I think I'm going to be launching the drone from.

And as soon as I do that, you'll see it starts to do these rows. Each one of those rows is based on the-- so you specify what drone you have, and what camera it has, and it automatically calculates. OK, at 197 feet, I'm going to be able to see this much of the earth. And then it automatically identifies, based on ways that this data is processed, how many rows you need. And it does this in real time.

So the thing you can come in and change is your altitude. So sometimes you have clearance up to 400 feet. Sometimes you have clearance just to 100 feet. So as I change this altitude, you'll see that these rows start to spread out, because you're able to see more of the Earth with each pass. So I'll lower that back down, and the other thing you can do is adjust what's called overlap.

So I mentioned, hey, you click on a pixel, and it shows you how it triangulated that. So theoretically, there's a sweet spot of about 70% to 80% overlap in photos that allows these software packages to go in and calculate the most accurate solution. So you're able to go in and adjust those. And all of those things have pretty major impacts on your flight time, which is usually a concern.

And in this particular software, it does support multiple mission batteries-- or sorry, multiple battery missions. And so this white pass right here is the first battery. And then after it turns to this darker gray, that's the second battery. So it's telling me I going to need two batteries based on the fact that we're flying the Matrice 210, and I've told it the batteries are going to last this long based on past performance.

And it's going to tell me-- when I had that magnifying glass open, it showed me how big each pixel represented. And it was showing 0.5 inches. And that's all based on ideal conditions. So if a cloud came over, it would increase my mission time automatically and slow the drone down, so that way there was never more smear. It didn't cover more distance than half an inch with that pixel.

So in Pix4D-- I'm not going to spend a ton of time on Pix4D. It's its own animal. But I did want to focus on some export settings, because if somebody is doing these missions for you, and they're processing the data, the thing we need to focus on is this ground sample distance. And that is how often it's going through and calculating a data point based on that flight that you did.

And the other thing that we need is this orthomosaic, and I mentioned that a little bit earlier. And that's that image that looks like a satellite photo. We need to make sure that merge tiles is on. Otherwise, it'll export each little patch in that quilt as its own image, and that's a pain to combine in AutoCAD or Photoshop. Not that I've manually had to do that, but you know, if I did, I could tell you it was painful.

The other thing we want to focus on is, on the additional outputs tab, there's a grid DSM. And so AutoCAD Civil 3D is not particularly great at importing these DEM files, which is how this software normally wants to kick it out. It's effectively an image, and its colors correspond to an elevation. Civil does not like that, and it doesn't treat it like a normal surface. But if we export this grid DSM, it's almost the same data-- a little bit different. But this is something that can come into Civil 3D.

Before we spend a whole bunch of time playing around with Pix, there's an option to export contour lines. And if you're going into Civil 3D or any kind of engineering solution, that is painfully slow to use. It more than quadruples the amount of data, depending on what contour distance you set. So this is actually a more accurate form of the data, and then Civil 3D can interpolate those contour lines as needed.

So don't use the contour lines. We want to use this grid DSM. And it'll save out an XYZ file that will come into Civil 3D.

And I mentioned this already, but after you create your digital model from Pix, we went out, threw it on a rover, and walked the site, and made sure that it was indeed accurate. So that way we could make sure that our calibrated measurements were indeed calibrated and correct to the site.

So now we're going to jump into some Civil 3D. Does anybody have any other questions on all the content we just covered?

AUDIENCE: So when you said--

ERIC CYLWIK: Yeah, go ahead.

AUDIENCE: When you say it was accurate, what was your criteria [INAUDIBLE]?

ERIC CYLWIK: So less-- it had to be more-- it had to be within a tenth of the actual surface for elevation-wise, and I think we were like 0.05 horizontally. But that was extremely important since we're tracking quantities and doing a volumetric calculation for it.

AUDIENCE: Have you got any experience with [INAUDIBLE] photo in Pix4D?

ERIC CYLWIK: So when we were doing this, ReCap Photo was just getting to the point where it could support GCP. So I haven't done any direct comparisons. Sorry. The question was, have you compared ReCap Photo to Pix4D outputs? We have not done any measured and quantifiable outputs like that, but it would be an interesting exercise.

Let me duplicate my screen so you guys can see it. Any other questions? Yeah?

AUDIENCE: [INAUDIBLE] has the zigzagging track from-- and I think when I was experimenting with that, it didn't have the double mapping, where it's crisscrossing, and Pix4D does.

ERIC CYLWIK: Yeah.

AUDIENCE: Do you notice and difference between double mapping it, and just using the zigzag?

ERIC CYLWIK: So the-- yes. So the question is, do you notice any difference? So the grid pattern that I showed is just a single pass. It'd go through and zigzag. There's another option where you can do a cross hatch, and it'll rotate that same pattern 90 degrees and run that the other way on the site. And the other thing-- one of the options that I didn't show is you can also change it so that instead of pointing straight down, it points at a 70 degree angle.

So the crisscross pattern is typically used when you have the camera at that 70 degree, or oblique, angle. And that's used so that way you can capture the different sides of 3D surfaces. So if you're flying over a building, and you want to be able to show the exterior of the building, that's the kind of mission that you would fly. You would fly an oblique perspective one.

The nadir images, where you just have the camera pointing straight down and you usually just do the single pass, tend to produce better results when you're looking at just 3D surface information. So if what you're trying to capture is more flat than vertical, you'll usually just need a single grid pattern. You don't need to do the crisscross. And you also want the camera straight down.

Does that answer your question?

AUDIENCE: Yes.

ERIC CYLWIK: Perfect.

AUDIENCE: I was going to set my survey-- well, I've set the survey to [INAUDIBLE] before and laid the target down. And then they would survey the target, so they knew what the elevation for the target would be. Is that overkill, or would you do that?

ERIC CYLWIK: So it depends on what kind of accuracy you're looking for. The GCPs that I mentioned are laying down a visible target. It's a cross pattern that you can see visibly from the drone. Those four permanent ones were around the site. They were brass in the street. And then we had just targets that you can lay down. And then, while they were flying the drone, the other person would run out and do a five-minute capture on a GPS to be able to identify the center of those targets.

But that was just driven by the fact that we needed to do volumetric calculations. Normally, we don't do that level of detail. I've got time for one more question.

AUDIENCE: Also, so the export that you're doing out of that, was there a lot of [INAUDIBLE]?

ERIC CYLWIK: Great, great question. So the question was, what's the file size of the export from the Pix4D program?

This was about a 60 acre site. And the export at a 50 centimeter grid was about 36 megabytes. So the data set itself isn't particularly large. And I'll actually address one of the issues with it here in just a moment. But it's more about how you look at that data in Civil 3D than it is about the actual size of the data set.

So jumping into Civil. I am a dork, so I know the first thing I need to do is come in here and change my drawing settings. And I have memorized my coordinate system like all good humans. Uh-oh. So I'm in Arizona central in this case. So that just sets it up so that when it downloads the Pix data, it knows where it's located in the world.

I'm now going to right click on Point Groups and go to New. So to address your question about the data size, when you import a point into Civil 3D, it wants to represent it with a live annotation tag. Hey, here's the marker, and here's the data associated with this. At 36 megabytes worth of points, you don't want that. It takes an hour to import it, and then every time you edit your drawing, it's got to go through and redo all those. Pain.

So to deal with that, I'm creating a new point group, and I'm telling it that the point style is none and the point label style is none. And that will make it a lot faster to work with this data set, because ultimately, I'm interested in the surface and not the point anyway.

I'm now going to go to the Insert tab, and I'm going to say points from file. And then I'm going to hit this plus to select my file. And this is that format-- it's in this XYZ format. You have a few other options here, but we're going to dump it out of Pix. And it's in that 50 centimeter grid pattern, so every 50 centimeters there's an elevation point.

I'm going to hit open. And then Civil 3D is automatically going to limit my list here to just the ones that it thinks meet that pattern. I'm going to click on XYZ RGB here. And I'm doing this part right now this, two seconds here, just so I can get through the demo without it taking forever.

But I'm going to come in here to Manage Formats, XYZ RGB comma delimited. And I'm going to say Copy it, and I'm going to say sample every 12 points. And so what this is going to do is it's going to read line 1 and create that in the model. And then it's going to skip down to line 12 and read that into the models.

So it's going to skip a lot of the data. So I'm just effectively weeding the points in a very crude way. But it'll cut my import time down to two minutes instead of 20. So I'm going to hit OK. I'm going to call this AU2018. OK. So now I've got this AU2018 type.

And then it shows me my preview down here of my easting northing point elevation and then color. And I'm going to say add the points, and I'm going to choose that group I just created that has that no style look to it. So I'm going to hit OK. And while that imports, I'm going to jump back over to PowerPoint. And we're going to sneak ahead.

So the purpose of getting all this data into Civil is so that way we can produce something that we can turn around to the owner and confidently say, I've moved this much dirt. I know that they say they brought in 15 trucks from the landfill, but what does that actually mean? How full was each truck? How compacted was that material?

And so over here on the left, we're looking at that again, that orthomosaic. You can see some of the stitching from where the photos were taken along here on the side. And then up here on the top right, there's an Excel spreadsheet that we produced, again, bi-weekly. It had a super small snapshot of the entire site, and then we identified those six categories that we were tracking for quantities.

And then we hooked that up to some cells below. So that way, as we recorded more quantities, it would automatically fill out that bar chart. Taking the super abstract, weird idea of counting trucks, and which quantities are going where, and we boiled it down to six bars in a bar chart. And then to back up that information, we also identified a cross section location, and then showed the owner what their site looked like on day 1.

Then on the first time that we flew after the start of construction, we updated that spreadsheet. We had our six bar charts that started to populate and show exactly what quantities we billed for. And then we have an image over here showing how much that site had changed. So day 1, and then we were maybe a month into the project at this point.

But you can see we came in and our scrapers needed to get down in the pit, so they hogged out a whole bunch of this dirt. And that corresponds exactly to what we see over here in our cut/fill map. So this is super tangible now. Something that's like, well, how many scraper truck did you do? Doesn't matter. Doesn't matter how many scraper loads I did. You're looking at this, and I have a quantity, and you can physically see what we've done out there to make progress happen on that job site.

So just skimming through some of these. We'll see the site evolve over here in the top left. And then on that right side, you'll see exactly where we were digging down to be able to get out trash, and also excavate a little bit for a parking structure that was going in place. And you'll see that story of us overrunning quantities, which normally is a pain.

But the owner's looking at the data in real time, and we're able to have those discussions. Hey, do you want us to keep on going with trash? The inspector told us we should. Or you can tell us to stop, and you can call your inspector and tell him to back off. Oh, no. Keep going. OK, so then you know that you're going to overrun on this quantity.

And so we're creating all of this information and bubbling it to the surface in a very, very tangible way. I mean, here we're starting to get foundations in for the actual building, and starting to level out some of that surface for the parking garage. And then we did a final. And so in between here, we pulled out a whole bunch of garbage. So over here we pulled out a whole bunch of garbage, and then brought up the rest of the site to be able to be at grade for that building and the parking structure.

So let me jump back over to Civil. Perfect. So it's finished importing the points. If I click on my point group, it's going to show me I have 72,000 points. And that's one out of 12 for a 50 centimeter grid, so just a ton of data. I am now going to right click on Surfaces.

So you'll notice it didn't update my view at all. And that's because I put it in the point group that doesn't show me any points. Yeah.

So I'm going to right click on Surfaces, click Create Surface. We will call this one AU2018. I'm going to hi OK. If I wanted to be crafty like Autodesk, we could call it AU2019. Release the product before the year. Let's do that.

So I'm an expand it, look at the definition of it. I'm going to right click on Point Groups and hit Add. And I'm going to click on Point Group 1 and press OK. And then again, it's going to look like nothing happened in my view. I'm going to do a Zoom Extents, and there's my surface. So that is straight out of Pix4D, through an XYZ file, and into Civil 3D at a 50 centimeter grid.

Yeah, question?

AUDIENCE: I'm just curious why you didn't just import a points [INAUDIBLE] surface. Because you've kind of got points upon a slope.

ERIC CYLWIK: Yeah. Yeah, so-- yes. This is the easiest way I've been able to get the data in and manipulate it in a way that I need to. And I don't want any visual-- I don't care about the points. I just want the data from the surface. So I don't--

AUDIENCE: Yeah. All the more reason to just import the points back to the surface.

ERIC CYLWIK: Yeah.

AUDIENCE: [INAUDIBLE]

ERIC CYLWIK: Yes. Yeah, so you could do that as well. Yeah, I like separating them a little bit, though. I like silos.