On Grid: Tools and Techniques to Place Reality Data in a Geographic Coordinate System

SETH KOTERBA: All right, guys. Thanks for coming in. My name is Seth Koterba. I'm one of the senior engineers on the ReCap team at Autodesk here. And my colleague here, Ramesh Sridharan, he's one of the senior engineers on the InfraWorks side to reality capture data, reality data.

Ramesh and I came into the company through a similar fashion. We both came in through acquisitions, as most people have in Autodesk actually these days. I started about four years ago as part of an acquisition called Allpoint Systems, and Romesh was maybe a year or two behind me with-- part of a Virtual Geomatics.

Today within the company, Ramesh and I are couple of the leaders with reality capture, reality data. We're kind of helping define where reality capture goes and the future of that. If you guys are interested in that or you have opinions about that, please come talk to us. We're happy to hear your feedback and get some input from you on those things. As far as this class, I want to talk a little bit about the motivation for it.

You know this topic about taking data and aligning it to a survey coordinate system, or a geographic coordinate system, is kind of a common subject. And we have a lot of questions about it from our users. We see it a lot on our forum posts. And it's something that we don't have a lot of documentation on. And it's actually kind of a concern-- confusing ways to try to do some of this stuff. So I wanted to make some clarification there. You know documentation-- I wouldn't say anyone who would accuse ReCap of having good documentation on anything, but these particular features are probably less-documented than others in general.

And there are some gaps in some of the workflows or some breakages. And I wanted to point out some of those and tell you the things that we're working on, but tell you how to get around them today. And then obviously there's a growing interest in reality data, reality capture, in general, but this topic of getting it geolocated is definitely on the rise as well, especially with starting to do more UAV work, photos in general in the area.

As far as the outline for today, I'm going to start off talking about some definitions and some of the basic stuff within ReCap to help us set us up for some workflow examples. I have a number of examples of workflows. I'll go through all that stuff. Might be a little boring, a little dry, but then Ramesh is going to come in for the fireworks later and show us all the cool stuff that's going on in InfraWorks and Civil 3D.

All right, so I'm going to-- the obligatory key learning objectives. I'm not going to go through all of these. The point is we're going to learn some stuff today. Talk first about some definitions. I want to talk about the difference between structured and unstructured. This is some terminology that we use on our team. I'm not sure how prevalent it is within the rest of the industry, but we talk about it a lot. And we differentiate the two of these.

We talk about structured data is that traditional, tripod-based terrestrial scanners. They are structured because all the measurements come from a single origin, right? They're all taken from the center of the mirror on the tripod there. And the data is ordered in rows and columns. It allow us to create these nice spherical views. And we take advantage of some of that information in some of the algorithms and tools that we use in the application.

In contrast to that is unstructured data. And this is data that doesn't have really any organization. It's just a bunch of xyz points. And .las is an example of that and also unified data. So if you've taken a bunch of structured data and then combined that together into a point cloud, you've basically thrown away that kind of information about the individual measurements of the data, where it came from. So it becomes unstructured in that sense.

I'll differentiate the two of those at a few different points in this presentation. The other thing I want to talk about is coordinate systems. There are a couple of different types of coordinate systems. You may be familiar with a lot of these, but geographic coordinate systems is a coordinate system that provides a coordinate for every location on the Earth. And these tend to be spherical coordinate systems-- your typical latitude, longitude, altitude types of coordinate systems.

Because it is a spherical coordinate system, it makes it kind of hard to work with. If we're doing design work, we prefer to have Cartesian coordinate systems. So that's why we use a lot of projected coordinate systems. So a projected coordinate system is one where we're taking a planar approximation of the surface of the Earth in a small area.

So you have the xyz coordinate system, or sometimes we talk about it as easting northing. And you know there's lots of examples of those. The US State Plane is one of the popular ones around here. So we have a planar approximation for every state, or it might be multiple ones per state. So those are the main coordinate systems that we'll talk about. There's one other one that I'll mention quickly is Earth-centered Earth-fixed.

You don't see this one a lot, but it is used from time to time. And this is both a geographic coordinate system and a Cartesian coordinate system. Because we're taking an xyz coordinate system and slapping it into the center of the Earth, the origin is at the center. The z-axis is roughly along the line of the axis of rotation, and then the x-y plane is roughly in the equatorial plane.

So I'll mention that one again a bit later. But it isn't commonly used in design work, just because wherever you happen to be on the surface of the Earth, z isn't necessarily up, unless you're working at the poles for some reason. The other thing I'll mention is targets and survey. When you're working with geographic data, trying to get onto to-- reality data-- trying to get into a coordinate system, you're generally going to want to have some survey data. This isn't always the case, but in many cases, you're going to need to use survey.

And the way that that's often done is by using targets. So you're checkerboards, you're spheres, any kind of a uniquely identified object. And then a surveyor is going to be able to identify those objects with their tools and then create some kind of text file or something for you to give you those targets and the locations in that coordinate system.

And then within your reality data, then you're going to go in and you're going to extract those same targets that the surveyor gave you coordinates for. And then enter that information shows and examples of that in a bit. But that's one of the key pieces to a lot of the workflows that we're going to be talking about. All right, then I wanted to talk also about some basics of registration and some of the tools that are used in there, because those are going to come up a bit later on in the workflow examples.

So let me first show some of these. Let's see-- jump into ReCap here and show you a quick demonstration of manual registration. So I have here a data set-- I have just three scans that I've pulled in-- and you notice that down on the bottom here, I have the option to click on auto-register, but if you hover over that, there's a couple of other options. So there's skip registration.

So if you've done your registration outside of ReCap or another product, you can just directly skip that step. Or do manual registration. So I'm going to show you the manual registration process quickly. First step is to just grab one of those scans and have that be your anchor scan. And then you'll have a split screen here.

And what we're trying to do is just find common points between the two scans. So I'm going to click on this sign here and here. And then I'm going to click on this cubicle wall here and here, and do that same thing in this scan here and here. Then you'll notice that once I've done that, we get a kind of a preview of how those scans are going to come together, and it gives you some sense of that.

But if you-- let's say you had missed and got caught the wall behind there, you can see that initial alignment doesn't look very good, and it's telling us that it's a poor match. So let me go ahead and move that back in. Our match looks good again now. And then we can merge that in. And then when that's done, it'll give us some information about the metrics about how that aligned. I'm not going to go into a lot of details about that if. You're interested in that, you can come see me at the booth or after the presentation.

I'll explain in more detail how those things work. The other thing that you can do is do that same kind of process from a top-down perspective. Sometimes this is easier in certain environments, where you can kind of see the outlines of the building. And then it's easier to click on things, so I can click on this pillar or maybe this pillar up here. Same thing over here and over there. And it would bring the scans together similar to doing the three clicks in real view. Now you can do just two clicks in the top-down view, because we're making some assumptions about what is up, and we just want to align it rotationally.

So that's the general process for manual registration. If I go back into the import menu here, or import tab, I'm just going to show you real quick what happens with auto registration. So if I click on auto register, it's telling me I have already got some registrations. So it's just going to blow that away and start from scratch With auto registration, what we're doing is we're going through and we're looking at pairs of scans and trying to match pairs of scans, and then build up a network off of those pairs of scans, and ultimately bundle-adjust that whole network.

And in this case, the auto registration was able to find all three scans and grouped them all together. If you're dealing with larger projects, you might find that you'll get a group of five scans, you get a group of 10 scans there, or maybe one or two that it didn't find a connection between. And then what you would do is use these manual registration tools to be able to stitch those groups together.

And there's some tips and tricks about how to get better results of the auto registration, depending on the way you scan stuff. And again, come see me afterwards or at the booth and I can give you some hints on that. So that's the main manual and automatic registration tools. I'm going to show you some information about target registration as well.

Target registration is something that we've had in ReCap for a while now, and that allows you to actually do registration without doing cloud-to-cloud, just only using targets themselves. I'm not sure how well-documented that is or how many people know about it. But what I've done here is I've already done a cloud-to-cloud registration on these three scans.

And that's one of the starting point to actually doing the target registration is you actually have to do the cloud-to-cloud first. And you know there are plenty of examples where you might not get very good results using cloud-to-cloud, and these generally are more natural terrains, outside, kind of large open spaces and stuff. And what you'd want to do in those cases is potentially add some targets to your data to be able to do that.

And so just to show you quickly how that's done, we have two options here for extracting checkerboards and spheres. And you can click on that, and it'll pull that out. Actually let me just show you that. If you just click in the general area, it'll snap to that center point. And then you can label it-- 39054. And then the other thing to note is that because we've already done that cloud-to-cloud registration first, once you extract one of these targets-- it's actually in world space and knows where those other targets are from the other scans-- so it'll actually pull those out for you automatically-- not pull them out, but label them automatically.

So this particular target, I've already pulled that out in another scan, and so it knows which target I'm pulling out. So I don't have to re-number that one. And then you'll notice that it's starting to turn green over here. So if you've got three targets per scan that have been matched between different scans, it'll turn green and let you know that that's matched.

So I'm gonna go ahead and jump through a couple more scans and finish pulling out these targets real quick, and then show you some of the tools that can be used after that. So this one is the one I just labeled, so that one gets pulled out. I think this one, I've already pulled out, and then one more here.

All right. Now they've all turned green. And at this point, we're still using the cloud-to-cloud solution. The targets are right now not being used to do any of the registration. Actually, you can go into this reporting menu, and you'll see the metrics for a cloud-to-cloud. But you can click down here and see your target data, and you can see some information about the different targets.

What I'm going to do now is show you the target registration. So if you've got all your scans green here, you can hover over this icon and come down to this feature called enhanced registration. If we do that, basically what we're doing is we're blowing away our cloud-to-cloud solution and only using the targets. Now, if we come into this reporting page, you'll see that all of our cloud-to-cloud results are gone now.

And we come in here into the target, and you'll see that now we have different values for our-- RMS noise. And you can go in, you can look at things differently, so you can look at all your targets. And if one of these maybe had higher noise, you could go in and look at the connections and maybe filter by that. Let's say we want to filter by 37917, you could see all your scans that had this. And maybe there's one scan where we mislabeled the points, or grabbed the wrong target and had a higher RMS, you can go in and identify them through that process.

So that's how the target registration works. The other thing that I'll show you quickly is that if you wanted to add survey here, what you would do is you click on this little eyedropper and say make a survey point. And then you could either enter your coordinates directly here, or you could pull in a text file that has all that data. And that will populate this list. And then you can go through and select from this drop-down the target name that corresponds to this target.

And it will automatically pull in the registration information for you here. All right, so that's start registration. Last thing that I want to show you, which we'll use in a minute again, is unstructured registration. And this is a new feature that we added in 3.1-- ReCap version 3.1, which came out maybe a month or so ago. What I have here is I have a group of seven structured scans that I've registered already. And then I've got one unstructured scan. And this scan actually it's not really a scan, it's a point cloud that came from UAV photo reconstruction. And we've downloaded the point cloud from ReCap 360 photo service.

And what we can do here now is find correspondences between the scans similarly like we did in the manual registration. So I'm going to click here and click here. Click here, click here, and then I need one more over here, and here, and here. Perfect. One sec.

You want to be a little bit more careful on this unstructured registration. All right, I think I got this roughly in the right place. There we go. All right, so I'm going to go ahead and merge that in.

And the metrics here-- the metrics that we use for structured data don't work very well for unstructured data. So you'll notice that the results look quite a bit different-- quite a bit worse than what you would get from structured data. And that's something where we're still working on trying to improve that. But if we go ahead and-- now we've got everything registered, we can index and launch and can go in, and we can do some inspection of that. Quickly here. You'll see the advantage of having the combination of the laser data with the UAV data.

So if I turn that one off, you can see with just the structured scans alone, now you can get a much more complete picture with the combination of them. All right, let me just do one real quick. Show you a cross-section here of that data. For the alignment-- you'll notice that the photo data tends to drift a little bit on some of the edges. It doesn't do a perfect job actually.

You can see kind of in this section right here-- I don't know how well that's coming across on the monitor-- but where the photo project doesn't do a very good job in some certain areas. But it does match up pretty well on more flat surfaces and stuff. So it just can give you an example of how unstructured registration works.

All right, so that was all the-- some of the desktop prerequisite stuff.

AUDIENCE: [INAUDIBLE]

SETH KOTERBA: So this particular data set came from a UAV that had GPS data. So it was scaled based off of the GPS data. Let me talk about that real quick. Photo reconstruction is another piece of the portfolio that we have in the ReCap portfolio. And this is for doing photo reconstruction of-- primarily for UAV data, but you can do it for other objects as well.

But the way that this process works is if you have a bunch of photos from the UAV, you can drag them and drop them in to the photo service. This is recap360.autodesk.com. Generally, you're able to just click go, and it's going to process that and send you an email with result. Sometimes you may want to add some correspondences between the scans, and this is showing you how you do that.

You find some natural features between the two, and click on those, and make some correspondences between them. This gives the algorithm some hints about how the photos might fit together. Once you've added some correspondences to the scan, then you can go in and set some parameters on what you want to get out of the data, name the project, and so forth. And kick that off for processing. And generally you get a result back-- depending on the size of your project-- within a couple hours or a half day or something. And you'll get an email.

You'll notice there's a couple of options for quality of preview. There's always a free option. If you want to do the ultra, that costs five cloud credits. And when you do the ultra, there'll be some options for you to have more formats that you can generate. And one of those formats here you'll see is the RCS. So this is the ReCap point cloud file that you get with that.

So those are the two desktop registration, photo registration. One other thing that I'll mention quickly is the import settings-- the advanced import settings in ReCap. Something that's often not discovered by our users-- they don't realize it's even there, it's kind of hidden a little bit there-- is when you're importing data there's this filtering tab that's the kind of default that comes up. And then if you click on the advanced one, you'll see that there some coordinate system options there.

And in many cases, you're going to want to just set your current and target to be the same. I'll talk a little bit more about that later in the presentation about when you might want to have those be different. But this is generally-- in most cases, those will probably set the same. So just to point that out. And I think that's all the basic stuff that I want to show within ReCap. And now we can kind of put those pieces and tools together when we're talking about some workflows.

All right, so the first workflow that I want to talk about is where we have a photo project that has survey and also a laser project that has survey. In this example we're going to do, is kind of have parallel tracks and ultimately bring that data together within InfraWorks. In this case, so we'll take our photo data and our survey data, run it through ReCap 360, and then create an RCS file. And then on the laser side, we'll take our structured data, along with the survey, and create an RCP-- RCS-- and bring that in together.

So let's start first with the website, the photo project side. When you're adding these correspondences like I showed before in that video, one of the things that you can do is-- there's this little eyedropper thing that you can click on. It pops up, and it and gives you the option to give coordinates for that particular target, or that particular correspondence. There are three different options of what you want to enter that as. It can be in lat-long altitude. It can be ECEF-- this Earth-centered Earth-fixed coordinate system that I mentioned.

And then lastly you can use xyz. And this is what we would use for a projected coordinate system, like a state plane, for example. You'll notice that we don't give you the ability right now to actually specify which state plan, or which projected coordinate system, you're in. And we'll have to do that later in InfraWorks. That's something that we're working to remedy in the future.

So just a quick example of how you might do that. So this is a particular data set. We're just-- marked some spots on the ground. You can click those and survey those in from the surveyor and find those again in the photo project. And then if you've got your survey data in a comma-delimited format, like it is here, you can see on the side-- you can actually just copy that string and paste it right into the x field within this survey field.

Pop it up and just enters all that, so one little tip there to make it faster to enter this information into the cloud. So then you've got your photo project. You process that and you're getting your RCS files down from that. So let's jump over to the desktop side and talk about the laser project. Here, we're going to extract your targets, like I'd shown before. And then you're going to click on make that survey. You enter your survey targets either using a text file, like I mentioned, or manually entering them in these fields here.

All right, so let me jump into InfraWorks and talk a little bit about how to bring that data in. Let's jump over to InfraWorks. So I'm going to make a new project here called workflow 1 and then pull up some data and bring that in. [INAUDIBLE] projects. I'm gonna start with this project and bring this in. So you see-- you'll notice when I drag that in, it comes up. It's actually already got the coordinate system defined here.

And this happens because when I imported the structured data, I set, in the advanced settings tab, the coordinate system at that import stage. So in order for that metadata to actually be in the RCS file so that it will come in-- or the RCP file for it to come automatically into InfraWorks like this, you have to set that up front. It has to be known at the time of index.

And I'll show you real quick what happens if you don't do that. I'm just gonna move this one. You see what comes up is-- you have some additional settings here. So you actually have to go in and set that. And this is what we'll have to do for our photo project as well. I recall I mentioned that you don't have a way to specify which projected coordinate system you're specifying your ground control points in. So let me bring this one in. Show you real quick that-- the alignment here with some aerial imagery just to prove that it's actually in the right point. Pull that in. Why is that not [INAUDIBLE]? Is that what it is? OK. Am I not on the conference?

Oh well. You trust me right? All right, so anyways, that-- so let me bring in both the different projects here, so I'm going to pull in the laser project. And then let me pull in the photo project. And here you're gonna have to specify what the coordinate system is, because we couldn't specify that at the time of indexing in 360-- ReCap 360.

So there we go, you can see the two coming in together. And we have our alignment, so that is what I wanted to show you for workflow 1. All right, let's go to some other workflows here. So let's talk about another workflow where we're dealing with photo project that doesn't have any surveys. So it just has to the GPS data. In this case, we're just taking your data right into ReCap 360-- photos for 3D service.

And then we're generating an RCS. This is one of those examples where I was talking about broken workflows. So there's actually some tricks that are involved in actually getting this to come in the right way, particularly if the coordinate system information isn't being transferred into the RCS file properly. So this is a workflow that we're hoping to fix probably in the next six months or so. But if you need to use it today, or when you go home next week, I'm going to show you how to make that happen.

So when you've done your processing with your photo data in ReCap 360, autodesk.com, you're going to want to click on these three little dots that you'll see underneath-- at the corner of the project. And once you've done that, it's going to give you the option to download your data or click on this a360 folder. That's going to take you into a360.autodesk.com.

And this is actually giving you access to all the derivatives that came out of the process-- the photo process. And what you're going to want to grab is your RCP and download that RCP. So what makes this workflow even more confusing is that this RCP isn't actually the same as the laser RCP that you get from desktop. This is actually a photo RCP. They have the same extension-- historical reasons for that. But this particular file is actually a text file. It's an XML file you can open in a text editor.

And what you'll need to do in this case is you actually go in and find the tag that's called export. And under that, you're going to find your GIS data. And when you're bringing that project into InfraWorks, you actually need to copy in the latitude, longitude, altitude information into InfraWorks to get the results that you want-- to have the data located in the right area.

AUDIENCE: How do you reconcile the difference between the m and the z [INAUDIBLE]?

SETH KOTERBA: Say that again.

AUDIENCE: How do you reconcile the difference between [INAUDIBLE] altitude [INAUDIBLE]

SETH KOTERBA: Yeah. There's a separate data in there, I guess for that. But so the data is actually in a Cartesian coordinate system, which is the xy.

RAMESH SRIDHARAN: The data-- [INAUDIBLE] for.

SETH KOTERBA: Yeah, does that.

AUDIENCE: I did not.

RAMESH SRIDHARAN: So z-- [INAUDIBLE] z [INAUDIBLE] That's why here. So this is basically to position [INAUDIBLE] Make sense?

SETH KOTERBA: Yes, this is just an offset. This is just an offset to locate your coordinate, or your data within there. All right, so let me add to that workflow a little bit. And let's say you had a laser project that didn't have survey data that you wanted to align to your photo project.

So there's-- this isn't a real high-precision workflow. You're not going to get like millimeter precision or anything like that. But there are some tricks you can do to actually get your laser data to essentially use the photo data as survey to pull that in. So the way that this would work is you actually take your RCS file from your photo project and bring that into ReCap.

And you can actually find some natural features within that data and create a note for that. And I'll show an image of that in a minute. And then you can create a copy of these coordinates and then essentially use that like synthetic survey for your laser data. So find those same natural features in your structured data and apply survey coordinates to that. And then here I'll show you an example of that.

So here's for a photo point cloud. And I've just clicked on this graffiti. I found an a that I can identify again in the laser data. And I've created a note here. And you can see that it provides the coordinates for that if I just hover over that little circle there. And then I can-- there's actually a way if you click on that little-- on this little icon right here, it allows you to copy that to your clipboard.

And then you can paste that into a text file or directly into the fields in ReCap over here. But essentially what you do is go into your structured data. Then in your real view data, and you find that same natural feature. You would use not a checkerboard or sphere target in this case, but just a manual target like you were doing the three-point click in the manual registration.

And then create a survey data or make it a survey point and enter that information in there. And in this way, you actually can bring your laser data into the same coordinate system that your photo point cloud is. Now remember, when we added that photo point cloud into ReCap, or into InfraWorks, you had to enter this offset. So you actually have to enter that same offset again for the structured data when you're bringing that into the InfraWorks. Yes?

AUDIENCE: So the xy coordinates that you entered in is based off of a ground control point that you had? Or--

SETH KOTERBA: So there is no ground control in this case. Actually the UAV only had GPS data. So it's being located using the GPS data.

AUDIENCE: OK.

SETH KOTERBA: And now we're going in, and we're taking some samples off of that and applying that to the laser data. We're basically creating our own survey data off of the photo data.

AUDIENCE: I guess my question is, how does it know what your point of reference is if it doesn't have a ground control point?

SETH KOTERBA: Well, that's where the GPS data comes in. So when it's creating-- when the photo point cloud is being processed, the GPS information is actually stored in the images. And so when it does that photo reconstruction, it has a GPS coordinate for all 180, or 250, or whatever photos that you have. And once you've created that model, it's doing a best fit onto that GPS data. So we know exactly where all the relative points are within that point cloud. I don't know if that made sense or not. Yep?

AUDIENCE: You could have used photo targets on the ground, [INAUDIBLE] But in this case, you didn't.

SETH KOTERBA: Exactly. Yeah in this case, we just collected some data out in a field and didn't happen to have any survey data or anything. It was just using that GPS data,

AUDIENCE: [INAUDIBLE]

SETH KOTERBA: Yes. So there's three different ways you can make it-- the sphere, a checkerboard, or these are basically the manual clicks. So the clicks that you would do in manual registration, those points can actually be used to survey, so they're just a manual survey point. It's just one that you've added, it's not--

AUDIENCE: [INAUDIBLE]

SETH KOTERBA: So after you've done-- let's say your manual registration or auto registration or whatever in ReCap-- then you can go back and you can-- well you can do it while you're doing that as well. But generally I like to do all the registration manually first, and then I come back in and I apply the survey to that fully registered project after the fact. And then it just rigidly aligns that.

All right. I'm going to save some time for Ramesh here, so I want to jump into workflow 3. This is where-- an example where we have a photo project that doesn't have survey, but we have a laser project that does have survey. In this case, the workflow will look something like this. So we'll have photos that-- no information about them, no GPS, no survey data, but our laser data, we do have survey data. And we're going to bring that RCS file into ReCap and use that unstructured registration I talked about earlier.

So basically register our photos, download our RCS, do our laser scans, register those together. And then we're going to combine those two, bringing that photo RCS and combine them with unstructured registration. And then use that survey data that we had for our laser points and bring that in. And then ultimately you can just export out your unified product and all of your points, photo and laser, are going to be all combined into one. And then you can bring those into InfraWorks.

So let me show an example of that. Make a new one. [INAUDIBLE] Do you know what this is, Ramesh?

AUDIENCE: No Internet?

[SIDE CONVERSATION]

SETH KOTERBA: Oh I see why-- because I was not saving that little--

[SIDE CONVERSATION]

SETH KOTERBA: I don't know if that's gonna start up or not, but basically what I was going to show you is that when you drag that point cloud in, it's actually not-- it doesn't have the coordinate system information in it. So actually in that case, I did set the coordinate system up front, the current target coordinate system to be the correct state plane when I brought the laser data in. But because the photo project itself didn't have this state plane specified for it, when you brought that state plane-- or when you brought that photo point cloud in-- its metadata didn't have the coordinate system on there.

So when you're doing the unification, in order for that many data to transfer out to the unified RCS file, all the individual scans within it have to all have the same metadata in order for that to get transferred on. And so all the laser scans did, but the point cloud from the photo project didn't. And so none of that metadata gets transferred in. So that's something to point out is that you've got to have that metadata in all your-- for your unified have it.

All right. Last thing I wanted show was-- talk a little bit more about unstructured data. So the LAS/LAZ files. We have support for both LAS and LAZ files, starting with our most recent release of ReCap. We've always had LAS, but LAZ is something new. And the reason I wanted to talk about these coordinate systems in the advanced scan settings. So LAS files, generally they have metadata in them already about which coordinate system your data is in.

And generally the data comes from a service provider, and they've done the work to put that data onto a particular coordinate system. When we load those files into ReCap, we actually read that metadata. And we can enter that information automatically into the field that's called current here. Current is basically used for the coordinate system that your data is in. The raw data is on disk. And target is the coordinate system you actually want to work in.

And there's a number of cases where those might not be the same. For example, we know that LAS files sometimes have the wrong coordinate system specified as the metadata in it. So let's say you had a coordinate-- you had your LAS file, and it said it was in state plane, whatever with, in feet. But it was actually in meters. Somebody made a mistake and put the wrong thing in there, right? So in that case, you would want to actually-- so let's say-- it was actually in meters, right?

So we would want to put meters here even though the header said feet, it was actually meters, so we'd want on to say that the current is in meters. And then if we wanted to work in feet, we'd have to set our target to be feet. So those two things would be different there. Another example might be if you had data that's in lat-long as a geographic coordinate system. ReCap doesn't support working in geographic coordinate systems. It's only Cartesian coordinates systems.

So you actually set your current to be-- WGS84, for and then your target would be something like a state plan like UTM or maybe even ECF. So there's a couple of examples there. There are a lot of other times where these two might not be the same. So you just have to make sure that you're setting your current to be whatever your data is in now, and then your target to what you want to be working in. And then it does that conversion for you. OK.

So let's jump over to Romesh and show you all the cool things you can do with it once you've actually got the data into the right coordinate system.

RAMESH SRIDHARAN: So while this is coming up, which is really impressive I'm pretty sure you guys understand that the-- it adds more value to the point cloud data, especially when it comes to-- should we make it in presentation mode here so I am good to go.

So that gives you a lot of choices. Now we have a UAV data. When you process with just the GPS, it comes with a GPS-based elevation, which is usually WGS84, and you can do something with it. If you want more particular elevation model like, you said you, can put the control points, you can fix it so that it goes to that elevation-- [INAUDIBLE] elevation mode-- and that particular data when it takes it from there.

And then on top of that, now you can take the GPS data, or the UAV data, merge it with the terrestrial scan or the mobile scans or something, makes your data more powerful. Bringing the data together is one of the big bottleneck right now in the industry. Everybody can just work in silos. Bringing them together and using it for the processing makes your point cloud, your sensors, and our software more powerful. That's exactly where this is going.

So once you have the point cloud data, now you merged everything, now you have a comprehensive point cloud, what are you going to do with it is what I'm going to talk about. So if [INAUDIBLE] my presentation before, there are a few slides I'm repeating. Just bear with me. But I really want you guys to understand the impact of the processing that InfraWorks can do it. So it's kind of intentional. The end-to-end workflow. That's one thing that's exactly what we started doing in InfraWorks. So when it comes to point cloud, I know [INAUDIBLE] can take it, it has some surface extraction, then ReCap always does a great job on everything.

But when it comes to the customer, it looks like a multiple project products, and I need to bring the point cloud data, I need to come up with my own workflow to do it. It's a same thing. It's actually worse than other products-- actually competitors' product. So we wanted to come up with a tool that makes this point cloud processing as much simple as possible. So that user wont first of all shy away from using it just because of big data and things like that. And second of all, you can take best out of your data and do something with that.

And that's one thing I personally very interested in. So the end-to-end solution is what we are trying to do here. And it doesn't matter what point cloud it is. We are coming to a state that is in Autodesk, we are coming to a stage where it really shouldn't matter, from user perspective, how you collect your point cloud, where it is coming from. You should be focusing on what you want to extract out of it, and we should be focusing on the tools to deliver the information to you. That's exactly what we're doing. That's what this slide is about.

Collection is getting much, much easier, much better. Now the registration, Seth did everything great job. [INAUDIBLE] registrations available. And the processing, we're working on. IT'S hard to follow Seth, but we're working our best. So points to models. So this is something I tried to say this in this conference multiple times. It's creating a model, and a model has a maybe a negative connotation for any design people here, but model is a model. If you take it as a point, [INAUDIBLE] a design, [INAUDIBLE] point.

But the idea here is, point cloud have a lot of cool stuff. And how to convert them-- imagining you import a point cloud data, and it creates a 3D model for you from the point cloud data, so now we have a realistic model you can work with either for modeling purposes or design purposes, or feature extraction purposes-- anything. That's what we are right after. So that's why this points to models makes a big difference. We hear a lot of echo, or is it OK? I should shut up? OK.

So the extraction. So how we are doing it. Just to give you background information, I can always throw a tool and say, it does this. Go and use it. But we want to give you a little bit more information on how we're doing and why they are doing it so that we can get a good feedback from you and you will also appreciate what we are working on.

So the extraction is first thing we are doing is creates-- it extracts something that is like a sense from the point cloud data. Something that makes a modeling and the design more powerful. I am using the word irrespective very loosely. Irrespective that domain [INAUDIBLE] point cloud data, there's always something essence in the plant cloud, like a skeleton, you can take it out.

That's what we're doing in the first one. Once you do that-- this is the beauty of it-- once you do that, the further downstream information process is much, much more simpler. Most of the time I think we don't even apply it now because nobody crosses this threshold in any other part. You just go and manually extract some lines at a point. We think that's the end of it. No. That's not even a beginning of the point cloud data. You can do much more with it.

So this is what we're focusing on. This is the main importance. What are the value proposition you could get out of the point cloud data as we go through all these things? So this [INAUDIBLE], list orderly, it's just a step number one. Very powerful step. When you use it, you will see it. So what is-- the derivatives.

I just want to make it clear. I'm going to use the mobile [INAUDIBLE] data as an example, but I'll play the concept of [INAUDIBLE] applicable to everything. High-resolution data, back of the truck, obviously is not that great here. But you see there's a point cloud data. At least, I've seen it, I see it day in and day out. So you see, it's a much rich information cloud. And I see a lot of stuff here. Ideally I would like to convert everything into the model. That's what we are right after.