LiDAR vs photogrammetry with Skydio 3D Scan
Get a sneak peek into the results from tests comparing drone photogrammetry to LiDAR.
Learn about the accuracy, application, cost, and operational complexity of these technologies especially for mapping and surveying. The session is designed to provide a clear, data-driven perspective of LiDAR and photogrammetry in drone programs as well as debunking common myths.
Watch Ted Strazimiri, CEO of SkyDeploy, and Travis Sachs, COO of 3DS Technologies, along with Skydio to learn:
- In-depth insights into the accuracy comparisons between drone-based photogrammetry and LiDAR. -Practical knowledge on the best applications for each technology, considering factors like cost and operational complexity.
- A clear distinction between myths and realities in using these technologies for effective mapping and surveying.
Learn more about Skydio 3D Scan.
- We actually, um, wanna start with a question to
- the audience. We wanna get a sense of, um, you know, who we have in the room today
- live. And so, um, again, I think,
- um, if you guys are following, and gals are following me on, on LinkedIn,
- you saw
- this, this post a couple of days ago.
- Which team are you in, LiDAR, photogrammetry, or both?
- Um, this has been a constant in my career.
- I don't know if y- you guys, Ted and Travis, um, uh, which team
- are you in? And this has put you on the spot,
- but why not?
- [laughs]
- It depends. It depends what you're gonna do with it.
- That's always the first question.
- [laughs]
- Yeah. Absolutely.
- I'm, I'm team both for sure. They're both super important in their own
- way,
- and very useful together.
- Yeah.
- Yeah. And this is, this is what I've learned in the last few
- years, is like it all
- comes down to what the customer needs, the project
- requirements. Uh, both technologies have their strengths,
- their weakne-
- their weaknesses. Um, you know, the, the use cases,
- um,
- can be, can be very different for both.
- Um,
- I'm just looking at the numbers here, and see how,
- uh, we still have, like, 40% to go. Come on, guys, go back to
- the screen. Give us a vote in there.
- [laughs] Um, but yeah, it's, uh...
- I was actually looking at an article that I wrote with the same title,
- "LiDAR, Photogrammetry, or Both?" in 2020, um, when the
- iPhone 12, uh, Pro was released. And I'm looking at you, Ted,
- 'cause I know that you had an, um, some...
- You, you did have some experience in, in Apple.
- Um, but like it's, it was for me, it was, like, just a,
- you know, such a disruption that you can, you can carry, like, a LiDAR on your
- phone. Like, I, I think, I think that's, you know, in the last 10 years we've moved
- from, um... When I, when I studied in college like 13,
- 14 years ago, a LiDAR was, like, super expensive.
- And nowadays it's just, yeah, much more affordable and
- accessible.
- Yeah.
- So-
- And that's great for everybody.
- Yeah. Exactly. And it brings... I think that's why this topic
- is so relevant,
- 'cause it makes this, like, you know, what should I use, is more accessible, so
- like they, they are both, uh, accessible to different projects.
- Um,
- cool. Let's have a look at what we have in the
- room
- today.
- Yeah.
- Oh, yay.
- [laughs]
- Did you expect this?
- I'm surprised by the photogrammetry number actually. That's higher than I expected.
- Yeah.
- [laughs] What are your thoughts, Travis? Did you expect this?
- Um,
- yeah, I'm surprised by the photogrammetry number
- too.
- I think maybe it's just,
- um,
- easier to access that, that technology.
- That's it.
- So maybe that's why.
- 100%.
- Mm.
- Absolutely.
- Yeah. Cool. Well, we have a
- team both winning. That's very exciting.
- Um, I'm gonna encourage everyone to use the
- chat box throughout the webinar.
- We... If you've been in, in a Ted and Alicia webinar before, we love,
- um, you know, engagement from the audience.
- So feel free to use the chat box, and we'll
- keep an eye on questions and comments as we go,
- as we move forwards.
- Um, so yeah, I think we're gonna, we're gonna start with the very basics.
- So for anyone that is an expert in LiDAR and/or
- photogrammetry, um,
- you know, uh, st- maybe, like, lay back [laughs] and chill for a few
- minutes.
- [laughs]
- We wanna just get everyone on the same page.
- So Travis, let's start with you. Um, you have a lot of experience on the
- LiDAR and laser scanning. Like, tell us, tell us, you know,
- in a nutshell-
- Yeah
- ... what this technology is about.
- Sure. Uh, yeah.
- So, so basically we're, we're gonna talk about terrestrial
- scanning for this, for
- this, uh, webinar. Um, and terrestrial scanners,
- um, there's two different kinds.
- Well, there's more than that, but we're, we're just gonna talk about the two
- different kinds. So there's time of flight scanning, and
- there's phase shift scanning. Um, the time of flight is more
- accurate, and it holds the accuracy over the full
- range of the, of the scanner.
- And basically how that works is it emits a beam of light.
- It waits for that light to return, and then measures the distance.
- Um, and then with that, some of the models can range up to, up to
- 1,000 meters of distance and hold that accuracy.
- Um, phase shift scanners, it emits a f-
- phase of light, um, and then the reflective light
- that comes back,
- it measures the time and the distance.
- Um, and then it gives you a much faster
- return.
- It's still, it's still highly accurate, but not quite as accurate as a
- time of flight scanner. And, um,
- uh, the further away the distance, then you kind of lose that accuracy.
- Um, and then each point will give you back,
- um, an X, Y, Z coordinate.
- And then if you're capturing pictures with that, it'll record that RGB value.
- Um,
- so when you're doing terrestrial scanners or scanning,
- you wanna know, first of all, what you're scanning, and then
- that will help you
- determine the setup positions. Because it's only line of sight,
- you wanna make sure that you're capturing
- all of the object that you're, that you're trying to capture.
- So you'll have to move your scanner into different positions, um, to create a
- full 3D point cloud, which is basically because you're measuring up
- to two million points a second. You, you know, you'll wanna make
- sure that, that
- all those points that are, that will create that 3D model will
- cover everything of, uh, you know, of what you're gonna do.
- Um-
- And, and, and this comes... well, goes back to, if, if I, if I can
- add here, 'cause I think we will, T- Ted will, will, uh, talk about this,
- um, I think in the photogrammetry, um, section.
- But, like, this comes down to good planning as well.
- Like, before you go on site-
- Yeah
- ... you'd really need to plan this in advance.
- Yeah. So you have to know, um-
- You know, what, you know, we always say, "What are you doing with the scan?" So you
- have to know what you're doing with the scan.
- You'll have to make sure that you have, you know,
- a clear line of sight to everything that you're trying to scan.
- Um-
- Mm-hmm
- ...
- uh, because you can only, um, capture what you can see,
- then, uh, you wanna make sure that you, that you're able to see what you need to.
- And then, and then-
- Mm-hmm
- ... so with the registration of that, so each, um,
- setup position
- you'll have a mass of, you know,
- 3D points, and then it, you know, they'll sort of be
- individual
- point clouds.
- Um, and then you'll wanna put them together.
- So there's different methods you can use for that.
- Um, you know,
- typically you'll wanna have some sort of control network that
- you capture with a, you know, total station or, or even if you just set
- up targets. Um,
- but also with these softwares now, they use lots of
- algorithms and they, you know,
- they'll use walls, they'll use, you know, fixed objects,
- and they'll try and calculate that registration, um,
- throughout the whole... So instead of just using single targets
- at, you know, so you might get 100 points with the
- out the cloud registration, it'll use all those millions of
- points and try and match those millions of points to the other millions of points,
- so it's a much tighter registration.
- And then, um, so w- with that example we just saw,
- um, that's two different s- scan locations.
- And then when you take a very thin slice, then you can
- s- see how,
- you know,
- how those registrations align to each other.
- Yeah. Thank you for that.
- Yeah.
- Sorry that I, [laughs] that I was advancing too fast.
- [laughs]
- Yeah, no, that's okay.
- Yeah, no, we can talk about this for hours
- and hours, so
- trying to share [laughs] all the relevant information in, in a
- short amount of time is, is challenging sometimes.
- [laughs]
- Absolutely.
- Yeah. So this, um, you know, again, the first question we always
- ask is, what do you want from the scan? What are you trying to capture?
- And that will, uh, determine the outcome 'cause you don't,
- you know,
- sometimes you don't wanna overscan because then you get
- too much data because you, you know, you know,
- we're capturing literally millions
- and millions of points a second. So, um, those data
- sets can get very heavy,
- you know,
- quickly. So we wanna make sure we do proper site
- planning.
- Uh, we wanna make sure we capture what we
- need,
- um,
- a little bit more than what we need, but not too much more than what we need,
- and
- you always wanna, you know, always wanna make sure you capture enough.
- Um, it's always better to need it and not have it than to have
- it, or to have it and not need it than need it
- and not have it 'cause
- then you-
- Mm-hmm
- ... you would have to return to site.
- Mm-hmm. And these comparisons, they're all from terrestrial laser scans,
- so static
- scans. With, with... I'm seeing some comments about, uh,
- SLAM-based scanners,
- mobile scanners. Um-
- Yeah
- ... anytime the scanner's moving, like, there's more, there's more to account for,
- right? So, uh, the, the, the true sort of reference
- comparison here is always gonna be re-, um, terrestrial laser scanning, uh,
- static scanning.
- Yeah.
- Um, so some of the sources of the errors
- that you can get is, you know, distance.
- As you, as I talked about with the phase shift
- scanners, as you move
- further away, you know, those,
- you start getting more noise, range noise, and things like that.
- Um,
- if your scanner moves during the scan,
- um, you know,
- some of the higher end scanners like the Leica P Series,
- you know, they have dual axis compensation so they
- can compensate for that.
- But, you know, the mid and lower range
- terrestrial scanners
- don't have that compensation, so you'll get, you know,
- you'll get errors in there.
- Um, you know, if there's any,
- you know, um, if there's anything with your
- s- system calibrations that's wrong or, um, not
- up to date,
- you know,
- um, it gives you false return. Um, some of the things that, you know,
- they're talking about the, um,
- they're talking about some of the mobile scanners, that data you can actually
- manipulate. Um, with the terrestrial scanners you can't
- manipulate that
- data. It's, you know-
- Mm
- ... one to one. It is what it is. Um, some of those mobile
- scanners you can really
- run into some big errors pretty quickly if you don't ha- you know, if you don't
- have a really tight control network.
- Mm-hmm.
- Mm. So what you capture is what you get basically, and you,
- there's very li-
- little,
- um,
- room for, you know, improving or changing or, that, that
- data in the post-processing.
- Yeah, exactly.
- Yeah.
- So in those cases then you would get the errors from if you don't have a
- proper registration, and that's, you know,
- where you put the two together.
- Mm-hmm.
- Um-
- That's one of the limitations for sure, but, like, when it comes to where lidar
- shines, um, and I, and I think this is, uh, Te-
- Te- Te- Tetras Travis where, where does lidar shine and where are the,
- you
- know, the use cases where it is definitely the go-to?
- Yeah. So,
- you know, some of the ones if you're doing really tight
- or complex environments-
- Mm-hmm
- ... um, dark areas for sure, like if you don't have the lighting.
- Um, if you have areas like the
- draft tube shown in the picture, um, you know, if you can't
- really send a
- drone in there or, um, the camera, um, because
- it's dark, because it's, you know, tight, awkward spaces,
- um, you know, sometimes the long range scanners can get
- places that,
- um, even the drones aren't able to.
- Um, and then if you're doing, like, a large-
- ... complex site and you need to hold that accuracy, like a whole, you know, a
- whole factory,
- um, it's,
- um, it's much easier to capture that data with a terrestrial scanner.
- So it'd be good for that application.
- Mm-hmm. Mm-hmm.
- Yeah. I think when I was at Bentley, I really...
- There was a significant number of, uh, infrastructure,
- um, owner operators that were looking into, you know, the
- hybrid coming from we have like a, you know, LIDAR
- laser scanning for very complex environments and
- adding the photogrammetry layer on top of that to create that
- hybrid model that was ultimately the best possible digital
- twin they could,
- they could work with.
- Um,
- so yeah, this is a very good use case or example of
- that, I believe.
- Yeah.
- Uh, yeah, and then, you know, as you can see in terms of
- the LIDAR,
- you know, the applications are endless.
- You can use it for,
- you know,
- many, many different applications.
- Mm-hmm.
- And that list is growing 'cause I think we're starting to see the benefits of, uh,
- of, of certain applications and, and how much easier this makes it.
- Yep. For sure. For sure.
- Let's, uh, maybe touch base on the cost here.
- Um, Travis, what, um,
- what, what co- what could you tell us about [laughs] the
- cost of-
- Yeah. So-
- ... uh, these type of projects?
- Yeah. So it ranges, right? You can get, um,
- a low range,
- um, terrestrial scanner, you know, in the tens of thousands of
- dollars, and then you can go all the way up to
- $250,000. And, you know, some of the mobile
- mappers, um, are in the millions.
- Um, you know, and
- the difference between those will be the range,
- the range noise, um, you know, angular accuracy,
- things like that. And, um,
- you know,
- the speed of the scanner, you know,
- um, all kinds of things. And then
- with the scanner you also have to take into consideration like the annual
- calibration costs, and then typically most scanners
- you have to purchase a registration software.
- There are free registration softwares out there, but to get really good
- results, um, then you'll wanna have
- one of the proprietary, uh, softwares.
- Yeah. And I think, again,
- we go back to, you know, the project requirements and,
- and the customer
- needs, right? Like I said, it depends on budget is a key one [laughs]
- when we're talking about which technology we're gonna be, you know,
- leveraging for
- specific projects, you know, what's the budget, um, you know, what
- are the final
- requirements in the deliverables that, that will, you know, help you
- decide is if do you really need, um,
- laser scan? Do you need... You know, what, what are... and, and what's the budget?
- Um-
- Yeah. Yeah
- ... so do you find that a lot? Do you find that... Yeah.
- Oh, yeah, and that's one of the things
- that we help our customers with 'cause, you
- know, if you just need a floor plan, then you don't need a
- $120,000 P-Series scanner, right?
- You know, a Matterport might be fine.
- Um-
- Mm-hmm
- ... you know, so there's... You know, it all depends on the application.
- We always ask, "What are you trying to do?" And we'll
- try and help our customers make the best decision.
- Um,
- a- a- and in a lot of cases, photogrammetry is great.
- You know, it's a great solution. So...
- Yeah. Absolutely. Um, let's move on to
- photogrammetry, Per. Um-
- Yeah.
- Let's start with the basics.
- Yeah. I think, uh, most of the audience probably has a fairly
- good understanding
- of photogrammetry just because it is so accessible and people
- are starting to
- really, uh, try it themselves, and that's really the best way to
- understand how it
- works. But fundamentally, it's measurements from photos.
- So, uh, in order to make those measurements,
- we need photos,
- and those photos need to be overlapping and captured, um, you know, in
- a, in a very, uh, consistent manner.
- Um, same distance from surface is very important.
- That, that kind of dictates your, uh, accuracy that you can achieve.
- And i- if that's w- the same across the entire, uh,
- subject area,
- then you have a, a much higher chance of, of getting, uh, the
- accuracy you desire. Uh, but, you know, any photo works.
- Uh, in this case, the, the, the data you're gonna see, um, the...
- All of the imagery was collected with, uh, the Skydio S2, S2
- Plus, um, using the 3D scan software that is on the drone
- itself a- and that allows it to take over
- and do all of
- the data collection autonomously.
- I just give it a, a, a section, a volume, um, and it just goes
- and collects. And both the laser scan and the photogrammetry
- was done on the same
- day for this asset.
- Um, so fun- after you have those photos, the
- photogrammetry software goes and matches them.
- Essentially, uh, the, the unknown is where was the camera when the photo
- was taken? Where was the drone, and w- how was the camera pointing, uh, when
- the photo was taken? So that's, that's done by, uh,
- creating first tie points.
- Those are sort of like, like fingerprints, like little signatures within each
- image, and then matching those tie points.
- And you can see here how those matches kind of correlate to each other.
- These are three images that are, uh, captured, uh, sequentially.
- Um, after that's done, we can register those cameras.
- So it's ve- it's very similar to how those, uh,
- laser scan stations are registered
- together, um, except we're registering images.
- Um, but how do we get 3D from 2D images?
- Well, through that process and the densification of those points and,
- and more triangulation essentially-
- Right
- ... uh, we can extract these depth maps from these
- images.
- So essentially a depth map is like a, it's still a 2D image file, but it
- has, you know, another dimension there that's usually,
- uh, shown
- in a, you know, what, what can be represented as this grayscale
- color and, and that
- is just the distance away from, uh, that, uh, that camera.
- Um-
- And, and similar to-
- From there
- ... with the laser scan, uh, Ted, I guess that's, you know, planning,
- uh, you know, your flights is essential, right?
- Like weather conditions-
- Yeah
- ... overlap. Like, th- these kind of elements that
- are gonna have an impact on your
- final deliverables, uh-
- Yeah
- ... in the quality of your final deliverables.
- Yeah. Exactly. And, and, you know, laser scanning is, is an
- instrument that is very finely calibrated, very, very accurate from
- the factory, and it, it doesn't really ... It doesn't fluctuate very much.
- But cameras fluctuate a lot, and there's many, you know,
- models of cameras, many,
- many lenses. All of those factors come into play.
- So it's really, uh, about controlling those, uh, unknowns
- as much as possible. Um, so having something that takes some of
- that w- work or some of that effort away from, uh, the user really
- helps. Um, and, you know-
- Mm-hmm
- ... from there you get to what you see here.
- So, um, the, the main, you know, two main derivatives that we see, uh,
- with, uh, photogrammetry are textured, uh, meshes which, you
- know, give you this photorealistic surface.
- Um, and then from that surface you can extract
- or sample points.
- Um, and some photogrammetry, uh, engines do this a little bit differently.
- They kind of skip the depth map, and they just make more tie points.
- In these examples that I'm gonna show I used Bentley
- ContextCapture, or iTwin as
- it's called now, um, and, uh, iTwin, uh, does that
- route where it goes depth map, m- uh, mesh, and then sub-samples the, uh,
- the, the points from that mesh. Um, so yeah, let's dive into the
- he- uh, the comparison. Um, here's the, the, the area that we scanned.
- You're seeing the laser scan, and you're seeing the point cloud, uh,
- from the
- photogrammetry overlaid on it after I've, uh, registered those two things together.
- So which one is the ...
- Which one is which?
- Yeah. So the laser scan is the one that's-
- Um-
- ... uh, brightly colored. Those, that's showing you the intensity of each
- return.
- So laser scanning-
- Mm-hmm
- ... can't show you color, but it can show you how, uh, you know, how
- reflective the surface that that pulse of light or that, uh, uh,
- phase of light
- bounced off from. And, and that kind of replicates color quite well.
- So you can make out certain details.
- Uh, you can't really use it for inspection,
- like bridge inspections or structural
- inspections, uh, but there is some detail there
- that allows you to sort of, you
- know, conceptualize what you're looking at
- and see the, the difference in materials
- as well.
- Mm-hmm.
- Yeah.
- Yeah. Uh, so for, uh-
- And I, I-
- ... this specific ... Yep.
- No, sorry. I was gonna say, I expect that that's already a, you know,
- for certain
- projects an advantage on photogrammetry versus LiDAR.
- Sure.
- Um, you know, you, you get, like after the, the flights
- and the processing you already get like a representation of your physical asset,
- which you do as well with LiDAR. Excuse me.
- But you don't have that color, um-
- Yeah
- ... in your final 3D model. Yeah.
- Yeah. The co- the color-
- You do have the option with LiDAR. Yeah.
- Yeah.
- You can add the photographs after-
- Mm-hmm
- ... and then each point will have the RGB
- value.
- Yeah.
- Um, but it's an extra step after the, after the data capture.
- Yeah.
- Mm-hmm.
- Yeah.
- Yeah.
- Exactly. Yeah.
- A- and so the RTC that we used here has, uh, c- color
- cameras on it that it al- aligns for you.
- So, um, you, you do get both. The, the, the main difference though is that
- that ca-
- that camera that's on the laser scanner, um, that's not really meant for
- inspection. It's fairly low resolution.
- Mm-hmm.
- And, um, it's really just there to help you
- colorize the point cloud, uh, to give
- it a more natural look and also to help you with registration,
- which is ...
- What you'll notice, there's some targets around this building
- that I set up, and,
- and, uh, using that camera built into the scanner
- was, uh, how these two
- things were merged, uh, in, uh, in the d- derivative that I was, uh,
- aiming for here. Um, so for, for this specific-
- Do you have a sense of the,
- the, the time of the capture? Sorry. Sorry, Ted.
- [laughs]
- Time of-
- Do you have a, a sense of the time that you spend on ... Yeah.
- For these two.
- Uh-
- With LiDAR versus this data.
- Oh, so how long?
- Mm-hmm.
- Yeah. The, so the lasers, the laser scanning moved along quite
- quickly.
- I think, um, Travis' employ-
- It was like-
- Yeah, like less than an hour
- ...
- yeah, it was more than an hour. Yeah.
- Yeah. Some- something like that. Laser scanning for small sites like this
- moves
- very quickly. Um, the photogrammetry-
- Mm-hmm
- ... was about two hours, so not significantly more.
- Um, but again, there isn't much area here.
- It's just a lot of, uh, you know, dense, uh, data collected in, in one
- spot.
- Um-
- Mm-hmm
- ...
- so let's, uh, take a look at how we're gonna compare
- these.
- Um, basically we're gonna use the terrestrial
- laser scan point cloud as the
- reference. You know, that's the ground truth.
- We're gonna assume that is the most accurate representation of
- that, uh, asset that
- we can get. Um, and then, uh, we're gonna do two different versions
- of the photogrammetry point cloud.
- So the first version is unconstrained, so meaning no controls.
- Um, really the only thing that was there, uh, is what the software was able to find
- in the images and match, and also those, uh, those stickers that I placed
- around the, the surface. And, and, and those help align everything.
- Um, it, it, it ... This asset actually didn't really need
- those stickers, but I put
- them in there because it really helps bring the laser scan
- and the photogrammetry
- data together in, in, uh, post-processing.
- Um, the other version of the point cloud
- that we're gonna compare to the laser
- scan is a scale constrained point cloud.
- So same data except now we're giving it one
- reference of scale. Um, and you'll see why that's important,
- and you'll see sort of what difference it makes.
- Um, here's the, the, the basically
- raw, unscaled photogrammetry point cloud
- besides
- the, uh, LiDAR point cloud. So you can see scale is,
- uh ... It has obvious errors, right?
- The, the house appears to be much bigger than
- what the terrestrial laser scan dimensions, uh, show us here.
- And let me just clarify 'cause I've seen a couple of comments.
- So, um, just to be clear, we, the one on the right
- is the one coming from, uh, photogrammetry.
- Mm-hmm.
- And that's unconstrained right now, but you, we will see another example
- which is
- we've introduced a scale constraint.
- Um, and then the one on the left is the one you captured with,
- um, laser scan.
- Yeah.
- And, uh, you could have added color to
- that one as well.
- Like there is an option to have like a colored- Okay, cool.
- Mm-hmm. Yep.
- And then there was a, um,
- uh, I, I'm not sure you've, you've added this into the example,
- but like there is
- a, there is a way that you could basically merge these two-
- Mm-hmm
- ... and create what Bentley Systems calls a
- hybrid model.
- Yeah.
- So you use the, the laser scan data-
- Mm-hmm
- ... and then the photogrammetry together to
- get the best possible 3D model.
- Yeah.
- Yes.
- Yeah. And I have a, a post on-
- Okay
- ... on LinkedIn that shows the, the final, uh, derivative of that.
- Um, but photogrammetry-
- Yeah. Make sense
- ... without controls is prone to the scaling error.
- This isn't, you know, this isn't a, a novel finding.
- This is very well known, so that's why we use controls when we can,
- um, to, to reduce the scaling error.
- So you can see sort of the section, uh, from the side of what that scaling
- error kind of appears to be. Um, you know, it's relatively accurate.
- The, the, the overall shape and, uh, you know, size is, you know,
- matches what the, the house looks like, it's just globally,
- uh, scaled up, um, due to that, uh, you know, unknown o- of,
- of exact dimensions from just camera images.
- Uh, to take a look at sort of what, like what the factor of
- that scaling in this
- example was from photogrammetry, um, I, I did some
- very, uh, basic, uh, point-to-point matching here.
- Uh, and it turns out that the scale factor of photogrammetry
- to, you know, to get
- it to match the, the LiDAR point cloud is about .87.
- So you can think of it as the photogrammetry was 13%, you know,
- scaled up than, than it would have, it should have been.
- Um,
- so that's version one.
- Which is not, which is not bad for certain use cases.
- Like that-
- Yeah
- ... would be good enough for-
- Yeah
- ... for what, what use cases will be, do you think that will be good enough
- then?
- Exactly. I mean, um, you know, for, for inspection, that, that could be fine
- depending on what you're looking for.
- If you're looking for, you know-
- Mm-hmm
- ... very general inspection a- a- information, um, you know,
- you're looking for
- where the cracks are, where the damage is, you know, a, a, an estimate of how much
- that area is, is, is co- is damaged, um, 13% isn't,
- you know, gonna make or break that inspection in, in a lot of cases.
- In other cases, it is more critical.
- So doing something like this is, is probably what you should attempt first,
- which is, um, giving it some reference, giving the, the photogrammetry
- software reference of what, you know, two distances within the scan,
- um, are. Um, so to do that in this test, um,
- I, I got that reference from the laser scan.
- So I had two targets, measured the distance between the
- two targets. Um, you can see here it's 3.8 meters roughly.
- Um, and I take this measurement, and I f- find those exact
- two spots in the, the photos. Um, and I tell it,
- "This is 3.8 meters, so from here on,
- use this, um, for the rest of the reconstruction." So
- that should bring everything,
- uh, back down to scale where we expect it to
- be.
- And this is available in most photogrammetry softwares-
- All of them
- ... uh, correct?
- Yeah.
- Yeah.
- Yeah, yeah.
- Scaling is such a basic thing. Again, it's just a rigid global
- scale. It's not like we're taking one side of the
- house and making it slightly
- larger than the rest. It's just one, uh, rigid adjustment.
- Um, and, and to do that, I mean, it helps to have a point of reference,
- a target,
- something to, you know, pick and say, "This is the center of the first point.
- This is the center of the second point." Um, in this case, I used, um, the
- AprilTags, which, which, you know, are very useful for many other things.
- Um, but you can see here that with the photos, you have a lot more
- precision in, in picking, um, that point that's right
- in the middle. Um, because from a just a point cloud, you,
- you
- know, you're, you're left with the points that you see.
- So, um, the intensity and the color that can be added to laser
- scanning helps f- you know, to find where that target is,
- but it, it, you know, there is some measurement error there.
- You know, where are you gonna click, and where is that exactly the center?
- And you can see in the photo, um, it, you know, it's very, very, uh, high
- resolution, so you can see that center very clearly and,
- and click within a few
- pixels of it.
- Yeah, it definitely provides more context as
- well. Yeah.
- Mm-hmm.
- So now taking those two scaled versions, so the
- laser, uh, as the reference, and then the photogrammetry that's,
- um, used that scale constraint, and doing a quick coarse alignment on
- those
- two point clouds, you can kind of see how they match.
- Uh, this is just a very coarse alignment.
- The laser is the green, and the photo, uh, the photogrammetry point
- cloud is the red. Um, so some places they're overlapping,
- some places one is under the other, e- et cetera.
- Um, and so we can get-- We can do a little bit more of a finer
- alignment, um,
- next and, um, and then we can compare exactly how these,
- uh, point
- clouds fit within each other. Um, so to do that fine alignment,
- um, all of this, you know, this is stuff you can do yourself.
- And if you have questions on how I did this,
- feel free to reach out to me directly
- after. I'm more than happy to show you how.
- But in a nutshell, we, we bring in the, the photogrammetry point cloud.
- We bring in the, uh, laser point cloud, use that as reference.
- Um, and then I disabled scale adjustment.
- So in this alignment process, it is not allowed to
- adjust the scale of the photogrammetry point cloud because I'm
- gonna, you know, I, I'm saying that I've already handled that
- scaling in the photogrammetry software.
- So, uh, you know, a, a very apples to apples comparison here
- straight out of the post-processing.
- And then here's the result of what it looks like after that final, uh, fine
- adjustments or fine alignment is done.
- So the, the point clouds are, uh, transformed, moved a little bit to
- fit as best as possible. And you can see, again, that area
- of where one is slightly, uh, out in front of the other.
- There's a little bit of deviation, so there's some distance between these two
- points in some areas. Some areas they overlap perfectly,
- and then other areas, the
- other one, you know, the, the photogrammetry point cloud is, uh,
- is, uh, in front of it.
- Um, let's take a, a finer look at this.
- So I, I'm taking a cross-section one centimeter
- wide, uh, profile,
- and I'm running down the side of the house here at the corner.
- And you can see the red is photogrammetry, the
- green is laser. Um, so these are very close.
- There are deviations though. There are areas where the
- photogrammetry gets it wrong, and most of the time that's because the
- material
- that we were modeling might have been metallic, might have had some, some
- reflection. Of course, reflections, anything that has, uh, that
- type of, of, of surface is gonna be very challenging for
- photogrammetry 'cause it
- just can't match points between multiple images.
- But on the other hand, the laser scan also has some areas where
- it just can't
- see. It's all line of sight.
- Mm.
- Um, so having, you know, uh, the two together allows you to
- completely cover the, the, all the surfaces and all the dimensions.
- Um, and so all of these measurements are, are metric by the way, so we're, we're
- looking at millimeter differences between the two.
- Again, the only thing that was, uh, done to the photogrammetry point
- cloud, uh, to, to get it to here is just a one scale constraint,
- right? So you don't need to put these scale
- constraints everywhere.
- You know, more is probably better, but in this case, just one is what
- got us to, you know, uh, four millimeters, seven millimeters,
- uh, six millimeters at the bottom distance
- between the two.
- Um, and the worst case is at the top there, at, uh,
- two centimeters,
- um, and that's because of that smoothing that photogrammetry can
- create.
- And you can see where that smoothing, what that looks like in the model.
- It's this sort of like melted, uh, eave trough, right?
- That, that's not really what it looks like.
- The LiDAR has no problem with that, but the photogrammetry will always have
- problems with that.
- Do you recall what's the GSD for the photogrammetry model?
- I don't recall. I think it was in the millimeter range.
- As you saw, this area was pretty tight, so we had to fly really close to
- it. So, uh, GSD, you know, the spatial resolution of the images,
- um, was quite high due to that fact that we were so close to it.
- And I think you mentioned already, but we used, um,
- you used ContextCapture for this specific project.
- Mm-hmm.
- But you, you are s-
- use... Skydio is, is... Skydio data is compatible with most photogrammetry
- platforms.
- Mm-hmm. Yep.
- Um, but, uh, I, I think you're quite familiar with both
- ContextCapture and
- RealityCapture. What are your go-to, uh, tools when it comes to
- photogrammetry?
- Well, well, if I wanna take both and create that hybrid scan, um, I
- prefer RealityCapture because it, it supports, um, structured point
- clouds, uh, in, in a way that allows me to very
- fine-tune that alignment between the photos and the laser scan.
- Uh, uh, but, you know, generally, um, it's what you're
- comfortable with and what you're doing.
- So with photogrammetry you have, you have lots of options.
- With laser scanning, you know, you, you use the software that, um,
- that scanner,
- you know, supports essentially. Um, at the end of the day,
- again, it's always what are you after? What's the deliverable?
- Um, and-
- Mm.
- ... ContextCapture is what I'm showing for the photogrammetry
- point cloud
- here. So the photogrammetry point cloud here has
- not been merged with LiDAR,
- or else it would look perfect. They would be, you know,
- a one-to-one match here.
- Um, but the, uh, you know, RealityCapture is great to combine them as
- well, and there's some other software there too.
- Uh, so here's another example. Um, again, those distances very close.
- Sub-centimeter, six millimeters at the bottom,
- four millimeters.
- Then you can see another one of those cases where the red points, the
- photogrammetry points, are kind of like melted.
- It's smoothed across, and it's not really ref- you know,
- showing the
- true geometry of that portion of the, of the house here.
- You can also see an area where the, the laser scan just didn't see anything.
- It was in a shadow, so there's no points from the laser
- scanner there.
- Um, and that's pretty obvi- uh, uh, pretty evident at the top.
- That's the roof.
- So the roof was on a pretty, uh, steep slope, uh, and due to that high
- angle of attack for the laser, you know, we're gonna get less points that come
- back and, and are registered on the scanner itself.
- But still enough points in this case to, uh, uh, to make a, you know, a
- surface out of it.
- Um, and now so just to, just to sort of show that I'm not
- cherry-picking locations, let's take a look at like a,
- a global comparison of
- the distance between these two point clouds.
- Um, and a, a wonderful software to do this,
- and it's called CloudCompare.
- It's, you know, it's completely open source,
- amazing software made specifically
- for this, uh, purpose. Um, and it's gonna compare the
- photogrammetry point cloud to the reference laser scan point cloud.
- Uh, and it's not just gonna compare to the
- nearest neighbor,
- um, because that distance may be incorrect,
- uh, but it is gonna use
- some local modeling to essentially create like a surface from that, um, uh,
- uh, laser scan, uh, and, and measure the, the, the absolute
- distance between the compared point and the laser scan surface.
- And so this is what it looks like. Uh, in this, in this color scale you're seeing
- the photogrammetry point cloud,
- but the colors that each point has inherited, the, the
- value there, uh, it shows us the distance of
- that point to the laser scan. Um, so
- anything in blue is basically zero.
- There's no distance, there's no delta between the
- photogrammetry point cloud
- and the laser point cloud. They're right on top of each other.
- Anything that's in the red is, uh, five millimeters or above.
- Um, there obviously is a lot of outliers here,
- things like glass and,
- and, and those surfaces that I showed you earlier.
- Um, I've, I've taken those out of account.
- I'm looking at anything within a centimeter of each other and how that
- relates globally to this, uh, to this, uh, scan.
- Um, so you can see very, very close. Like the ...
- You know, there are some areas where we're,
- we're seeing that centimeter distance
- between the two, uh, and those areas are kind of where you expect them to be.
- They're on that roof. Uh, you know, remember there's less laser points to
- compare
- to there anyways, uh, so it's, it's not, you know, directly indicative that
- there's, you know, systematic error or measurement error here.
- Uh, but the distribution is, is very nice.
- It really, um, does show that i- if, if this is done,
- um, you know, with, with even just one scale constraint and,
- and that's it,
- um, you can get really, really close. Um, is it ...
- You know, is it perfect? Absolutely not.
- There's so many variables in photogrammetry and photos to,
- to get this right.
- But is it good enough for a lot of use cases? I'd say probably.
- Um, and then here's another example
- And, and do you, do you see consistency across difference,
- um, like I, I
- know that this example was like a, a small house.
- Do you see consistency across different use cases?
- And, and I, I guess it's a good segue for the next,
- um-
- Yeah
- ... comparison with the bridge scan.
- Yeah. That's a perfectly timed question because
- consistency is key, right?
- When, when we're talking about measurements.
- Mm.
- So with this, with this example, you'll see that the results differ a
- little
- bit. And so to just to sort of introduce these
- two, um, data sets,
- the reference scan, again, this, this massive scan of this bridge.
- In fact, Travis told me that this bridge wasn't even what they
- were there to scan. It just happened to be nearby, and
- there was-
- Yeah
- ... enough, uh, data that I could use it for this.
- Mmm.
- So that's ... Yeah, that's an example of sometimes you get
- more data than what you need. Like, we were focused on the area around the
- bridge,
- but in that we captured the bridge, and it just so happened that, uh, for the
- purpose of this, it worked out well for comparison.
- Yeah. Yeah, yeah, I think I saw you post, uh, this bridge on, on LinkedIn.
- I was like, "Oh my God, I have a scan of that bridge too.
- Can I have that point cloud?"
- [laughs] Yeah.
- [laughs] 'Cause I really wanted to see this.
- Um, and again, my, my, uh, 3D scan, um,
- of this bridge w- we weren't doing the 3D scan for the
- purposes of measurements, um, we were doing the 3D scan for the purposes of
- inspection. So, uh, no controls were used, no scale
- controls even here. So this is nothing out of, uh, you know,
- referenced, uh, in the, in the photogrammetry.
- So let's see how it compares.
- Mm-hmm.
- So to compare it, I'm taking a, a small section off the, uh, the side
- elevation here of the bridge. Um, you know, there, there
- are some things that, um,
- wouldn't be a, a, a good one-to-one representation of the
- two point clouds.
- Um, for example, uh, in the next slide you'll see that,
- and you can see it here
- too, there's some shadows in the laser scan,
- right? So the ...
- It's line of sight. So, you know, Travis had the scanner
- on the ground, and the bridge is up there. It, it just can't see across there.
- If you really needed that, you would, you know, plan ahead
- and position the scanner
- in the appropriate location. Um, but sometimes that might not be possible.
- Um, and also there was a little bit of registration error
- here.
- So, um, I think because this wasn't really the
- intended target of the scan,
- this may have been overlooked. I don't think Travis made a mistake in,
- in his
- post-processing. I know what he ...
- I know he knows what he does very well.
- Um, but this is a pr- I think a pretty interesting example
- that shows that, you
- know, this isn't just, just put the scanner there, hit scan, and
- you're done. There is, you know, fine-tuning, there is some quality
- control that, that happens after the fact to get it just
- right.
- Uh, and photogrammetry isn't perfect either.
- Um, you know, even though-
- Mm
- ... the, the drone was able to see areas where the laser
- scanner may not have been
- able to see, we never put the drone through the bridge
- to capture these images and
- create the, the model. We're, we're very interested in the,
- the exterior,
- um, for this, for this quick scan. And this quick scan, it
- was about three hours.
- Um, and so, you know, it was at the end of the day.
- We had already done most of the inspection, and it
- was just sort of a test for us.
- So you can see sort of some of the, the deformities in that
- interior section.
- So when I bring the laser scan, and you can see the laser scan is the
- yellow points, and the photogrammetry points together,
- again, photogrammetry
- completely unconstrained, that scale error is still there, right?
- So again, not shocking. We know that photogrammetry has a hard time
- with scale. I didn't use scale constraints at all
- here, so we're looking at a
- direct output from both.
- And applying that same, um, uh, you know, alignment
- to see how different is the scale, you know, how far off
- we are, we're about 10%. So the scale factor is .9, so that
- means the photogrammetry point cloud needs to be shrunk, uh,
- by about 10% to, uh, to get it to,
- you know, a best fit with the laser scan. So a very rough best fit.
- Again, let's look at some sections. Uh, I picked this pier here.
- Um, you can see certain areas, it's right on the money.
- Other areas, there's a little bit of deviation.
- Um, you know, there's some registration error here to
- account for too, so not all
- of the points are valid. But it shows you that the overall shape
- and relative accuracy of the scan is pretty close,
- um, for, for, for, for photogrammetry.
- It, it gets it, you know, it gets it to what you w- would be
- very, very, very comfortable with to use in
- inspections or,
- or, or, you know, many use cases, uh, visualizations.
- You know, perfect for that. Um, when we take a,
- a deeper look though at the
- distances between those two points, um, in that, uh, in
- that section, you'll see that now we're, you know, much larger asset, much more
- distance to cover between the two, no controls at all, GPS with, uh,
- with, with, you know, issues that, that happen environmentally with photos
- there.
- So you can see that the distances are now in the one centimeter, two
- centimeters. Um, there is a, there is a bit more distance between the
- two point clouds, but still pretty damn close.
- And then for a sort of more forensic analysis, I chose one of the, you know,
- really flat elevations. Um, and this is the, near the
- abutment of the bridge. Um, and here I'm showing the distance between
- points again with the laser scan as the reference.
- Um, and then in that area on the right there,
- you'll see that there's some green,
- there's red and blue. So red means that the photogrammetry
- points were in front, you know, s- just slightly in front, and then
- blue means they were behind the laser scan.
- So these are, uh, signed distances on this cloud to cloud
- comparison. So overall, it's, it's, it, it, it fits quite
- well, but there are some areas where
- photogrammetry may have its own
- registration errors where those cameras were when, uh, when they were taken,
- right? So there is a little bit of deviation
- there.
- Um, and looking at the distribution of those
- distances across this area,
- you see that the distribution is quite tight. You know, it's, it's,
- it's within ...
- We have a, a, a, a mean of, um, you know, what is that? I can't even ...
- Looks like just under a, a millimeter. Very, very close for the most part.
- Um, but you know, you'll see at the- One centimeter mark, uh,
- we're in the 98th percentile of those points.
- So again,
- very close. Really depends on what you're using it
- for, but if you just need
- it for, you know, to have a model, you can,
- you can pretty much get away with,
- uh, with, with doing it this way, doing it very quickly and having
- that visual
- information as well. And then taking a look at the section, uh,
- as well,
- the whole profile of the elevation,
- we'll see here that it becomes fairly obvious that the
- photogrammetry point cloud is stretched vertically, right?
- So that's, that is, you know, not uncommon to see.
- Verti- the z-axis is always more difficult to,
- to get right.
- Um, in the center of the bridge, you can see that the two point clouds
- are well
- aligned, uh, but on the sides you can see that the, uh,
- the, the photogrammetry point cloud has less
- alignment, and specifically less
- alignment in the vertical.
- And again, just to look at some distances between the
- two points.
- We're- here, we're, we're looking at changes up to, uh,
- a centimeter,
- up to 10 centimeters in some cases.
- In the middle here where that alignment was the best, um, and
- that scale
- is sort of, uh, you know, less obvious, um, you can see that
- the, the, the distance between the two point clouds
- is in the two centimeter range
- as well.
- Cool.
- So overall conclusions-
- That, that was a lot of information.
- [laughs]
- Yeah.
- [laughs]
- Yeah. And so if anyone has any more specific-
- Let's
- ... questions on, like, how we do this and how to do it yourself maybe with your
- own data sets, like, reach out to me.
- I'll, I'll show you exactly the steps involved
- in, in, in creating these
- comparisons and these graphics.
- Yeah. Let's, let's go over some conclus- conclusions, uh, Ted and
- Travis. Um, and I think, I think my...
- Just to wrap up, uh, before we, we go over conclusions, like,
- what do you think
- y- I think, I think you both touched on a lot of the,
- the, you know, the
- weaknesses and the strengths of both technologies.
- Um, what do you think that is most, you know, the, the, the use
- cases, the best use cases for LiDAR, um, and photogrammetry?
- And I'm gonna start with you, Travis, and then we go to you, Ted, for
- photogrammetry. And then maybe if you can, um,
- think about what are the key benefits of a hybrid approach.
- Sure. Um, I think, uh, when it comes to LiDAR, um, especially
- terrestrial scanning, um, it's, it's the one you would
- wanna rely on if you want, like, if you wanna be,
- you know,
- certainly accurate. Um, it holds that accuracy better.
- Um, you know, in terms of, you know,
- you know, I've seen some comments about the mobile
- scanners.
- That's a whole different comparison, and I think, uh, in some cases,
- the photogrammetry beats the mobile scanners.
- Um, so with, in terms of terrestrial scanning, uh, if
- you're doing any reverse engineering, if you need, you know, to be certain
- about your measurements, then uh, you know, I would stick with the
- terrestrial scanning.
- Yeah. I agree.
- Cool.
- Yeah. The, the ter- the terrestrial scanning's gonna win the
- accuracy contest,
- like, every time, right?
- Yeah.
- It's, it's what it's, it's, it's what it's for.
- Um, I, I think one of the, the big things is just, you
- know, do you have the budget to have both?
- Um, or, you know, in most cases you kind of, you,
- you stick to
- one method over the other. Um, it may not be in your business, uh, case
- to, to, to, to dive into one more specifically.
- Um, and there's a lot to know.
- Mm.
- So, um, there's also the, the aspect of, you know, even if you have the
- hardware, you, you kinda need to know what you're doing
- with that hardware.
- Yeah. I always say, uh, especially terrestrial scanning,
- um, it can seem easy, especially with ones like these RTC 360s.
- You push a button and let it go. Um, it's easy until something goes wrong,
- and then it's not so easy anymore. So, you know, having that
- experience, you know, how to overcome some of those challenges
- is really
- crucial.
- Mm-hmm.
- Yeah.
- Um, both require a lot of attention.
- So y- you know-
- Yeah
- ... the, the, the, I think the, to get to very accurate
- photogrammetry, you need to know what you're doing.
- Um,
- less so for laser scanning. Um, there, there's, you know, the,
- just the fact that you don't, the, the, the distances between the scanner and the
- points, that's, that's it. There's no changing it after the fact.
- Um, that, that certainly takes a, a lot of, uh, unknowns out of
- the equation. Um, but really, it, it's also the size and
- shape of the structure, right? If you, as you saw,
- if you have a fairly small area,
- um, photogrammetry can get really, really close.
- Um, and using controls, using scale constraints, um,
- that only improves it. Uh, but if you have a really large area like
- that bridge,
- um, you have to take into consideration, like, is it, is it, is it feasible for
- me to photogrammetrically scan this entire bridge for the
- purposes of measurements? Um, that's where you might wanna take in
- someone
- that, uh, can help you, that, that has a laser scanner that can,
- you know,
- to, that can, that can help you and maybe even merge those two things
- together,
- uh, for the ultimate result.
- Ted, I have a question. Um-
- Mm-hmm
- ... in terms of the photogrammetry, if you're scanning, say, a windmill,
- and it has
- some slight movement, does the photogrammetry compensate for
- that?
- Um, it's, so it's a great question.
- Uh, in both cases I think movement is not ideal.
- [laughs] So-
- Yeah
- ... yeah, any change, uh, you know, a drastic change in the two
- surfaces, uh, will really affect, uh, your results.
- Um, now if it's a, you know, movement of the entire structure,
- um, then you're, you're probably gonna be okay, um,
- because you can adjust the camera position to, to match that.
- But it's, it-
- Yeah
- ... that introduces a whole, you know, that opens up a can of worms for both
- sides.
- So there's no magic bullet for that.
- [laughs]
- All right. I'm gonna-
- Yeah
- ... I'm gonna jump in. We don't have a lot of time left,
- and we have a ton of
- questions, um, that we're not gonna get to all of them.
- Um, but we're gonna get through s- a few of them.
- Just you can keep entering your questions because I will send these
- to, uh,
- the sponsor for follow-up via email.
- So don't worry if your question is not answered today,
- I'll make sure to send those
- over to Skyview.
- Mm-hmm.
- Um, so I wanna get through, try to get through a couple, okay?
- Um, so let's start here. Um,
- okay. I have a request from a client for
- archeological purposes
- with and without tree cana- canopy coverage.
- What is my best option?
- So the, I'm assuming that that relates to aerial laser scanning.
- Um, uh, that would be, you know, large area, you need to cover it.
- You have trees, you need something that can look through the trees.
- That would be laser, of course, right?
- Um, for the areas where there isn't tree can-
- uh, tree coverage, photogrammetry,
- uh, can work just as well. Um, but it's, it's that, uh, the combination
- of both that is gonna be really tricky for photogrammetry.
- Got it. Thank you. We have a, another question.
- How do you introduce LIDAR into infrastructures where there's no
- signal or physical access?
- N- no signal? Like, um, so we don't need a
- signal to use LIDAR, um, and physical access.
- As long as we can,
- like, as long as you can see what you're trying
- to scan, um, then you'll be able to
- capture it. Um, I'm not 100% sure I understand the
- question, but...
- Okay. We'll move on. How do you handle ground control points,
- and what do you recommend for GCPs on
- larger surveys?
- Uh, that's, that's a, that... There's a lot to unpack with how to put
- controls-
- Yeah
- ... and how to measure those controls.
- Um, so, you know, the traditional way is GPS, but as you
- saw, um, on the bridge example there, the question
- is: Where do you place, like, how do you get a GPS measurements for
- controls
- there? Um, so if it is, if it is a vertical structure like that bridge
- or that, that house, um, I would probably, if you do wanna use
- controls and, you know, known points, uh, coordinates, um, use something
- like a total station, right? That opens up another,
- a whole other instrument and
- a whole other, you know, set of expertise.
- Maybe, you know, bring a surveyor to do that for you.
- But it definitely is possible. Like, there's always a way to set controls,
- um, it's just a matter of doing it i, in, in a way that the measurement
- is gonna be accurate itself 'cause there could be error in that control as well if
- it's not done correctly.
- Yeah, exactly.
- And it, it, it goes back to the project requirements
- and the data
- requirements. Um, like-
- Yeah
- ... do you need absolute accuracy? Do you need relative accuracy?
- Like, what are you... Like a, like a, yeah, again, go back to
- what the customer needs. Yeah.
- Mm-hmm.
- Thank you. Uh, we have a question: What causes the
- photogrammetry to produce scale errors if all the
- images have georeferenced metadata?
- Uh, the answer to that is that the georeferenced
- metadata in the photogrammetry images, um, is not
- accurate, right? So in this case, we were using, uh, a SkyDio
- S2+. Uh, it does not have RTK capabilities.
- It's just using its, its, uh, GPS that it uses for navigation on top
- of its obstacle avoidance cameras.
- Um, so there's, you know, a lot of un- uncertainty about where that im-
- where that drone actually was in, in, in space when it was capturing that image.
- Um, so one thing I'm really excited for, uh, with X10
- is the addition of RTK. Um, that will get that
- positioning of the drone, that, that, you know, that prior position, uh,
- to be much more accurate, right, in that, in that centimeter, two-centimeter range.
- And anything you can do to have more accuracy going into the
- process will result in more accuracy, uh, at the end of the
- day. Uh, so that scaling issue should be really
- improved, uh, with the
- addition of, uh, of RTK or, you know, the differential
- GPS, uh, in a, in a camera system that's, that's flying.
- So, um, the, the other, the other side of it is the, the optics,
- right? It's, it's cameras, it's, it's...
- We're measuring from 2D images. Um, if the image
- is blurry, if the, you know, the lens is, is not great, if there's a
- lot of, you know, these aberrations, as we call them in, in photography, um,
- then that's gonna translate directly into errors, especially
- those scaling errors that you're seeing there.
- Thank you. Uh, we have a question: What is your go-to software for
- photogrammetry?
- Um, I think, [laughs] I think the one-
- [laughs]
- ... the go-to software is the, the software that you can, can
- get. Um, the RealityCapture-
- Yeah
- ... just went free.
- That's a good answer.
- That's huge, right? Like, RealityCapture-
- Yeah, RealityCapture went free. Interesting
- ... amazing software, completely free.
- If, if your business makes less-
- Wow
- ... than a million dollars a year, you don't need-
- Right
- ... to pay at all for it. Yeah.
- Wow.
- So-
- We... Let me, let me say that SkyDio is compatible, so SkyDio data is
- compatible with, um, all photogrammetry platforms
- that I know of.
- Um, and I think it depends, as Ted is saying, it depends on the
- customer needs, once again. Uh, not all the photogrammetry software
- are, you know, easy to use, intuitive, so it depends on the, you know, learning
- curve. Um, you know, how, what's the background of the people
- that are gonna be
- using it, the budgets. Um, you know, do you need very close,
- uh, final accuracy on the, sorry, accuracy on the final 3D mo- model
- versus, uh, you know, an intuitive or user-friendly workflow,
- cloud versus desktop. So there is a lot, a lot to unpack there as well.
- [laughs] Yeah.
- Um, all right. We have time for-
- Yeah. And a lot of the photogrammetry, so... Sorry.
- The, a lot of the photogrammetry software also
- is, uh, starting to make
- laser scanning as an input that, that it will-
- Mm
- ... work with and use very, you know, very effectively.
- And, and that's the best part about this, is that the, you know, ev- on both sides,
- they're, they're merging together, um, and it's much more accessible.
- We have time for one quick question.
- I think this one's for Ted, possibly.
- What kind of LIDAR scan would you recommend for use of surveying
- coastlines to document erosion, uh, challenges?
- Um, well, I, uh, I think there's two options you
- could take. I don't think terrestrial laser scanning
- is well suited for that, just
- because coastlines, to me, that sounds like it's a l- you know,
- long area, just
- a lot of distance to cover, and setting up these stations would take
- some time.
- Um-
- Yeah
- ... aerial laser scanning, you know, a lot of LIDAR payloads out there, a
- lot of LIDAR drones that are perfectly suited for that, um,
- are the way to go, uh, is the way to go there.
- Uh, and maybe even handheld if you wanna walk that coastline,
- depending on, you
- know, what, what the access is like.
- You know, both those, uh, types, aerial laser scanning
- and mobile laser scanning,
- it's not gonna give you the accuracy, you know, the millimeter accuracy
- that you're
- getting from terrestrial laser scans.
- But again, I'm, I'm, I'm thinking you probably don't need
- millimeter accuracy on
- a coastline survey like that.
Watch Time
56 Mins 18 Secs
Related Webinars
Pre-Incident Planning for FIFA: Turning Drone Data into 3D Operational Readiness
Watch webinar
Elevate crash scene documentation with Skydio and Faro
Watch webinar
Live Talk: Optimizing workflows with Skydio 3D Scan Indoor Capture
Watch webinar