Events Add an event Speakers Talks Collections
 
Marck Vaisman
Azure Data/AI Technical Specialist at Microsoft
  • Video
  • Table of contents
  • Video
Request Q&A
Request Q&A
Video
Processing LIDAR images for Forest Preservation
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Add to favorites
157
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About the talk

This talk will showcase how the USDA Forest Service is using LIDAR data to support large-scale forest management operations, conservation, and landscape-level ecosystem restoration. Marck provides a quick introduction to LIDAR and its benefits and using the lidR package to process images, and how using cloud technologies accelerates the process.

About speaker

Marck Vaisman
Azure Data/AI Technical Specialist at Microsoft

Marck is a Technical Solutions Professional at Microsoft and helps customers adopt the Azure platform and use it for Data Science, Advanced Analytics and Artificial Intelligence workloads. Marck designs data-driven computing solutions to help clients make better business decisions, recognize opportunities, experiment, gain insights, and solve difficult problems using large datasets and a combination of tools. His expertise lies in making data work for the problem at hand, drawing from experience in multiple industries including Internet, telecommunications, and high tech. Marck is an experienced R programmer and advocate. He founded Data Community DC, an organization that promotes Data Science and Analytics practitioners in the Washington DC Metro area. He teaches Big Data graduate level courses at Georgetown University and the George Washington University. He holds a B.S. in Mechanical Engineering from Boston University and an MBA from Vanderbilt University.

View the profile
Share

Jesus. You wanted the real early, I guess, progenitors of the data Community DC. The people he saw, what was going on New York, and he said, you know what, PC can use that? He helped build an amazing in DC, which we are very proud to partner with, to do this conference. They are in a bunch of really good people and we would like to give a big warm welcome to Mark Weisman as it comes up when I deleted the DC Community to give us his next talk. Please walk in the park. Yeah. Yeah. Hey

on the floor over there. Alright, have a good talk to. You might talk today is about a bunch of collaborative work that I've been a part of it for the last year and I'm really, really excited to talk about this. This is, this is a joint effort joint work between an academic government and Industry collaboration. So, several folks are involved in this. First, of course, is the US Forest Service, which is part of the USDA Department agriculture, the second kind of and they're all

related. But the second part of it is Florida A&M University and the sentence either which is the center for spatial ecology and restoration and of course Microsoft, so they're my customer. The u.s. Forest service is my customer. I do not work for Microsoft in a second, but this is basically a joint effort between Samu 4 a.m. You. Creed the center, which is a joint effort between the academic and government. And when I was going to give the talk, I I told Jared I said I can talk about this topic about using lighter images for Force management. I ain't ask permission because the work

really is is I am part involved in this but the work is really not. My work is really the work of these are folks, and they gave me permission to talk about this. So I'm really excited. Thank you again. And again, I'm really excited. So a big shout-out to Cesar at Family Center for special research. So, Jason Drake to Paul medley. Joseph Peter. He's probably online listening to the conference of a job and Jordan Vernon. So, I've been working with this team for the last year and they're doing some really, really cool work. A lot of time to talk about So I want to talk about

just give you a quick and turtle light are. So I didn't know much about lidar before working with these folks. I learned a little bit more in preparation for this talk. I'm going to tell you, I'm a show, you all the messages going to show you what they're doing and really, really. How they're using his light or images that far for really interesting analysis, and for bent Forest, management, conservation, and preservation, and all sorts of other things. I'm going to talk briefly about the lid, our package, which is with what they're using to process these light or images. And

then I'm also going to touch on a little bit on how we took a manual workflow that they had to move to San cloud and accelerate at that as well. so, oh, and I'm going to show you some food. I'm going to say some cool pictures of life for those that know me, you know, that I've been I've been part of the Sahara community and The Descent Community. Can't be here for a long time. Started dating me to deceive Rhonda, meetups, not as involved as today. I still looking for the conference and I wish, like I said to be in person

because I really, really love. I work for Microsoft and part of Microsoft Federal team. I work with Federal customers, that helping them migrate their workings of cloud. My specialty is indeed a science machine learning and artificial intelligence are in Tyson's Corner, Store all the open sores, but really, you know, helping our customers move, use our platform to do their work, you know, that I'm an orphan attic, right? Otherwise, I probably wouldn't be here. So, the water going to talk about today again, is the work itself, is Caesar's work at the ends. Are my man alone. Not necessarily

representative of anyone that I'm affiliated with her to say is for those actually. No, another fun fact is, I actually didn't when I started running the Meetup Group 10 years ago. I didn't know our and today quite earlier today. Actually. I saw a tweet in the I saw a thread on Twitter about people talking about why they had to learn our or what was the first reason. So for me, it was a G5, you know, just really that was like, the first thing I should learn but really the motivation was when I started running to meet up. I was really interested in are in the community by watching. Other

people, do really cool work with are like really motivated me to learn it as well. So that was ten years ago, of course A lot has happened in between So what is lidar? Later is similar to sonar and radar. We all know these terms, right? Basically. It's a sense that there's a signal, that got submitted gets received and it gets processed. So, you know, we know sonar right, you know, the like the beliefs of the clerks baseball-reference here, or are you sit back

and it does the processing. You kind of get a sense to what's out there? Well, guess what? Where are uses? Anyone care to guess? Stage anyone care to guess what light are uses? I'm sorry, drum roll. Laser. Yes, I'm being a total goofball. I know it's lighter means lights, action, and range. So basically what letter has many applications are, there's a lot of it out there, but in the context of what I'm going to talk about today, essentially is about selecting a Ariel lighter images from lighter units that are flown on Plants. So these are planes are drones

that fly over areas. So there's a, the ray of light is sweeping left and right, right. And it's picking out scouting the light off on collecting data at and so on and so forth. And that is the source of the date of that. That is that I'm going to show you today. The writer uses electromagnetic spectrum and uses the green light or the nearest Fred. And the reason why is because those are the ones that seem to reflect the best from vegetation and what you usually stay out in the field, right? At least let her know later is also used for example, in self-driving, cars

and a lot of a lot of new technology, right? And but, and it's the same principles just that smells it on a different thing on your car. Actually think that a lot of the cars that have a lot of the, the safety things or the stuff breaking at 8. I don't know what spells that use been pretty sure a lot of stuff to Prime cards. Use light are probably some of the commercial cars. Like the Subaru's. They have the two lenses probably using later. I'm not, I'm not sure. But anyway, but the point is that you can you can either admit later from the ground and made it from satellite or mines, which

is probably the most, the most common thing. What makes up the white lighter device for the light our system. It's the lighter unit itself, which is attached to the bottom of an aircraft or a drone. Right? And it's it's as if the plane is flying scanning at the meeting or later. Our way of life is going left, right? There's a GPS a right which pinpoints the position it helps also understand. I just asked why but also the elevation using GPS locations and then third, it also uses the inertial measurement unit because what happens is, you know of us a

plane flies at moves are also the light at because it's the rate of going back and forth left, right? There is an angle. So you need to know the angle of the planets in the angle of the light beam. So there's a lot of nuts, just the data gets collected back, but there's a lot of metadata that gets process to create all of these images. And what happens, right? When the ladies with the laser beam is going west to write, it's hitting on objects, but in the case of vegetation and probably other objects as well it penetrates so you actually get many responses at the same point. So when I talk

about the images a little bit, I'm going to talk about the response or multiple readings. And what you got. Looks something like that, so when you get all that data and suppose process, so there's also a computer and software that takes all the stuff and creates these images. What you get out of late. Our system is with call a point Cloud. So it's actually a set of 3D images. That is most specifically, essentially a multi-dimensional array where there's many different dimensions for of course, the the response that's why location in the measurement, like all of these things, get packaged

into this file type called when Cloud which is of the type of the the datafile itself is the pipe. Last so that's that's the last dance for me but it's a last while. So it's a point cloud and it's the last file at the standard. It's done by the American side of ranch and remote sensing. There's a good hopelink over here. This is just part of the specification. This was updated fairly recently. Okay, so that's really what later is and those are the images that that opposed to Caesar and other agencies are working

with, right? Lighter is actually, it is readily available. So, from what I understand is there is actually a lot of lighter data out there and a lot of government state, local governments, spend a lot of money collecting this data, for other purposes. So, and just to give you an idea, right? And I'm the statistics were right it to me. So, and in 2019, and 2020 the State of Florida, I think about, was it 23 million dollars for about 38,000 square? Miles of $600 per square mile to collect. There's another one here knows assistant, which is about $200 per

square mile. But the point is that they fly these drones or aircraft and they collect the lighter data for other reasons. Usually, for the purposes of what Freedom were called, digital elevation model, but the ancillary use is what I'm actually going to talk about, right? So the folks so it's easier and other teams have figured out and they said they're using. This data has been collected for other purposes for the purposes of Forest management. Like I said, they are there leveraging. This data has been already collected and get additional learning about this.

And one of the things that they're using is to calculate structural characteristics before I think was like, Forest Titan, canopy cover. Which, when you get that, it is a strong correlation to things about a forest about, you know, the amount of Timber the biomass, which is the amount of organic in your life, are the habitat quality. So they can use these measurements and also lay it with other data. I'm not going to talk about it. But you can also leave this with satellite data and these other things but ultimately it's about it's about understanding the metric about making decisions and

what they're doing with us, which is really cool. Right? As they're building their building hydrologic hydrology model their building Forest inventory. So they're trying to understand, you know, the evolution of forced overtime. They're they're trying to make strategic decisions, especially as it relates to event response. Natural event responsibly using the data in Friant post hurricane image of force analysis. For example, identify areas in the forest for maintenance and restoration and preservation. So there's just there's a lot of different things that you can do with a product

of these out of the state of Texas. And ultimately what happens right? You when you take these data sets and you leverage open-source Technologies and tools like our, for example, machine learning and the cloud. It gives you better accuracy for much less cost. Then been during the span of what, you know, it has to be before these lidar data, people have to go out to the field and collect the state of manually. So you can imagine you can't really scale a person to cover a broad area, you know, where I was in flight. You can cover a pretty broad area in a short amount of time. So therefore they

are Traders. But ultimately the long-term idea is to really leveraged the data fully integrate this into many processes and share the state across the agency's ETC to really build a a full-blown end-to-end process. So this is how you don't collect lighter data. I'm sorry, Joe. So, this was last year and I think it was October. I was down at the Apalachicola National Forest of the folks that these are family. They're based out of Tallahassee and we went out to the field. They were doing some testing of these fixed-wing drone, testing multiple vendors. This one. I believe

the sense fly to fix. One problem that we actually did not have a lighter sensor in it, but I just thought it was a bunch of bad actually do. There is a lighter sensor is not as sophisticated. The images. I'm going to show you come from different devices, but nonetheless, I wanted to show you that. I told you I was very simple pictures when you were called points out files. So this is what a point called looks like. So this is this is an image that coming from the D dry system and the Gatorade system is a really

high definition high-density lidar system. It gives you about 5,000 points per square meter, which is pretty done. Whereas other systems can give you anywhere from four to eight points per square, metre. I mean, it embarrassed with. This one is really Ultra ultra high definition. Of course, the higher, the definition, the bigger, the file, you know, the higher, the definition, the shorter, the or the smaller, the are you can cover on it on a given point for the detail. So there's different levels. Right of light are there sort of like lower low resolution Bay Area high-resolution lower

area? Sort of everywhere in between. This is an example of, I believe that this is part of the Apple. This is Western Apalachicola, and these images were taken right after her. Hurricane Michael in 2018, if I'm not mistaken, so I mean, you can't really see it here and these are just a couple of different examples of different points. And and that's all the date of that. Got selected as the beam gets thrown down. Do you get a response and all this data gets process. This is this, what you call the point Clapton, This is what is used for this ride data for Downstream analysis for building.

Are there other products that they're using for doing other kinds of Analytics. This one, then the temperature in particular, you can sort of see a little bit of of an area in in the blue, which is the are close to the ground. That's less populated. That's kind of an era where trees fell? So, this is this wizard of a flight passed right after the hurricane, Michael, which was in the Panhandle area. And this flight path was near the area, where the hurricane went through. So, just to give you an idea. So these are apoint clock file. This is a TIF file. So, what ends up happening is when you

collect all the data, you create our there's there's multiple levels here, but you can extract a lot of this. A lot of layers from the point cloud and also create a little bit creative files, which are you's down stream, but this these are multiplayer tips. So if you open an ammo dealer test and like a typical tip, you are like your computer, whatever image you are. You have, because it's multiple layer. You can't, you're not see anything. So, here we're actually seeing as the flight path, and if I come here, I take my mouse when I can hover it over this up. I can she can see it but

essentially what you see here is the whole flight path and and you see us a wide strip because that's the laser being that tanning left to right as the as a plane is moving. So this is part of that circular flight and the images that we saw that I showed you earlier. We're actually a part of this white but this is looking at the entirety of the image projected. I'm so single plane without color because it's just a bunch of data protecting the single play. Well, how did these get used? So they work on creating more cold raster products from these lighters when clouds. And when we say we're

after is basically means a grid. So it's Nebraskans is a two-dimensional but its petition by grid. And those are the kinds of images that are really use Downstream to produce further analysis. I'm going to show you some of the inputs and outputs the that are created. So for example, so this is a subset of that bigger flight that I just showed you. So this is just a hearing on a really small area just for the sake of illustration, but just because also image sizes are smaller and I can actually show you what a bunch of these things look like. So this is this is an aerial photo. This just

comes from satellite from the Ezra platform. Which of the spot forum. So when you overlay, the 2D point, this is just a rock to the point by but it's over late on the on the aerial image. Just so you get a chance to where things are. This is a 3D quick bite of that. Same area. Of course. It's not look, it's exceeded. The the point of view is not from directly from the top of her to come from side. But ultimately these are pointless but then we called the roster. So this is where you actually taking the front from the point card data, you can filter different bands and look at specific parts

of this. So this is looking at the canopy at the canopy height right side of like tree height. H nutso a blue is more open. Canopy red is denser camping? And if we go back to the satellite, right, you see that blue right there in this area over here or kind of towards the bottom, metal doors, left trees. Are you see that? It's because it's open the next one stressed 3 months. So that's kind of like this is canopy cover. So this is the height of a tree is this is actually the cover of the trees and you can see that the reds are more trees. More leaves, blue is less leaves and

so on and so forth, right? This is one version of it. This is so this is Can we cover this is another canopy cover here. Actually, this is shrubs. This is shrug covers. Tiredness labeled as light and blue is no or low shrub, cover, orange, or red meat sense. And then here, we're at usually combining all of these, these different layers into a single image. So the green is canopy cover. The, the blue is the density of shrubs mean lower trees. Right? And the red is the canopy. Canopy Heights of the really tall trees kind of where the exact dimensions

layer. Like in an integrated, right? This is a two-dimensional production and it read because every grid is this is resolution of 1 M. I believe 1 M2. So there are metrics are two. Populations are done herb bread for 3 Prince fail to produce these images, right? This is a sweet. Right? This is kind of a slice view of of the of the longitudinal sore lateral, sleep. So it's the same image. We're just looking at it from here. So you can see that you can do again. I'm not an apologist for an arborist store or something like that, but they're

actually doing a lot of being a lot of information, actual information from these images. They're using the package for that. So I'm just going to show you a couple of functions. Basically, what you do with the lid are so lit, are is a package. It's it's an active development. It seems to be the best pool out there for doing this kind of stuff in our particularly, but also, they tried a bunch of different workflows and it seems that this was the best. So you load the library library does use raster. It uses SP you need to eat. All is well for certain things, not for what I'm doing here

today, but the last file can't be there. Either part of a catalog or it can be a single file. So you can read in the catalog, which is this really the metadata. Remember he took his house is really bad. So that the big file, the gator I file that's about when I show you. That was about 8, GB. The other one that said was less about 500. Mags is a pretty large file. So when you can do a catalog and a catalog is basically a grid of images or a single image, but again, it's so here were reading in a single image. You can see that that image cover is about 4.0 Square km, there's 274 million

data points in that particular last file, but then we can substitute that you can clip it. Give it for the Nets and clip it and upset it. And here were looking at a smaller file, which is about 3636 thousand square meters. Much last seven million data points as opposed to a hundred and something something. So you can plot this directly with methods in our and get some of the images that I showed you and then you can post processing ultimately. What they've done is they've created a set of standard metrics are not inside, customer metrics. That basically reason the file and

produces all these sorts of metrics, four layer diaper. This is a list output and that's what they use to produce the downstream analysis. So, I need to wrap up. I know, so we, we actually migrated this work flow from local workstations, that was using data and hard drives, you know, a quarter machines with not a lot of memory into the clouds. So using large-scale machines with more CPUs, more memory. The data is already in the cloud the way through accelerated the the the process for being able to process. You know, how many many terabytes of data in the cloud. And that's one of the

things that I've actually been working on with them. But the team. So, you know, just the kind of wrap up. This really has a really huge impact because they're working on the scalable repeatable. Methodology that can be applied in again over many national forest is my other government agencies and also a builder. Warfel that folks that don't have made it the skills or the computer resources are going to be able to use and replicate. Like I said, this is no password for that. They've been able to find in terms of trying a bunch of different things and the raster image has those pretty

images are actually use it in place for machine learning models, that can be a whole talk on its own. And like I said, it really reduces the time to assess natural disaster in. Thank you. This talk was actually made with Sharingan. I've been using Sharingan a lot for the last year and building slides for a class. I guess, you know, I teach at Georgetown if I mention that I really like sharing and this is the thanks a lot guys. Mark mentioned, we were twinning because we're both wearing your DCR shirt, the blue DC, our shirts, but you noticed he was wearing a hoodie. It was

the Emily are. So we were twinning in more than one way and Mark was supposed to introduce me, but he didn't show up. So you didn't get his chance and he said for his fun fact that I should just take a jab at him. So, of course I had to be nice but marks an awesome guy. We love each other and he's doing really great stuff in the community. So our next speaker is

Cackle comments for the website

Buy this talk

Access to the talk “Processing LIDAR images for Forest Preservation”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free

Ticket

Get access to all videos “R Conference Government & Public Sector 2020”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Ticket

Interested in topic “IT & Technology”?

You might be interested in videos from this event

November 9 - 17, 2020
Online
50
244
future of ux, behavioral science, design engineering, design systems, design thinking process, new product, partnership, product design, the global experience summit 2020, ux research

Similar talks

Yvan Gauthier
Head of Artificial Intelligence Accelerator at National Research Council Canada
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Alfredo Corbett
Deputy Director at USAF ACC HQ A6-2
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
David Meza
Lead - People Analytics R&D, Sr. Data Scientist at NASA
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free

Buy this video

Video
Access to the talk “Processing LIDAR images for Forest Preservation”
Available
In cart
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free
Free

Conference Cast

With ConferenceCast.tv, you get access to our library of the world's best conference talks.

Conference Cast
946 conferences
37592 speakers
14370 hours of content