 This is your stage. Perfect. Thank you very much. Yeah, let's go ahead and get started I'm gonna click my timer here, so I don't lose track of time You for those of you that have stopped by the booth yesterday during the launch of the L2 at 3 o'clock or earlier today at 1030 we're gonna be covering a lot of the same aspects But overall just we can bring it both to the streaming audience and to be able to put it on a public stage We wanted to be able to talk from DJI's perspective about elevating precision Unlocking the power of drone lidar for accurate surveys. My name is Kyle Miller. I'm on the enterprise solutions engineering team I've been here for about two years and I've been able to start capturing data with the L2 about a month ago And I've been very very surprised and we'll show you why in here just in a second So if we look at the history of where DJI has been over 15 years ago for those of you that don't know about DJI We started off by making the main brains of the compute of a drone So the main flight control or components from there a little over a decade ago We started building building drones as all-in-one solutions that come with All the motors everything you don't need to assemble the drone You just are able to go and fly it recreationally Well once we saw that a lot of customers a lot of commercial employees Sorry, it's starting to slip. There's a lot of commercial clients that we're using these standard photo just nice Photography drones and putting them to work on a job site companies like Balfour baby Vinci BHP APG they started using, you know phantom twos phantom threes back on the day to get that new perspective That they weren't able to get from the ground So from there we knew that we had to put special focus on the commercial in the enterprise space We started an enterprise program in 2015 and in 2017. We really started putting focus specifically on the geospatial market So in 2017 we released the matrice 210 that had onboard RTK Also a fun fact in 2017 We released the first all-in-one drone with a mechanical shutter with the phantom 4 pro looking at accuracy when we start mapping in 2018 the next year we converted that phantom 4 pro into a phantom 4 RTK So now you have a small all-in-one solution that's tagging at with absolute accuracy rather than relative And then to be able to process the date process that data in 2019 We came out with DJI Terra that's going to be utilizing your photogrammetric 2d maps and 3d models in 2020 we really had a breakthrough year with the new matrice 300 series and then the subsequent payloads the P1 the full-frame camera That's great for photogrammetry and then the L1 our very first four-day for a Into LIDAR So that was early late 2020 early 2021 when we came out with those payloads In 2022 we came out with the most budget-friendly mapping drone with the Mavic 3 Enterprise series We did a nice little revamp on the matrice 300 with the 350 earlier this year But it's been the first time that we have released any sort of LIDAR product in the past two and a half years Since the L1 and now we're able to introduce the L2 The main Aspect of the L2 our main slogan is powerful range and elevated precision So when we dive into that we're going to highlight on six different aspects of how the L2 has Changed from our last generation of it with the L1 First off we're getting higher much higher Precision within our data capture. We're going to talk on the next slide Why we're getting that data accuracy improvement? But now we're seeing on the horizontal accuracy at five centimeters Vertical accuracy of four centimeters and if you read the fine print that was flying at a hundred and fifty meters off the canopy As well as capturing at 15 meters per second So we're flying faster and we're flying higher while maintaining a better overall accuracy compared to the L1 We also have Exceptional efficiency moving on so now that we can fly faster now that we can fly higher and maintain the same sampling distance Or a sampling rate as well as the accuracy now we can just overall capture more on a single battery So our efficiencies start to improve We do have in enhanced workflow improvements So with live view being able to see that point cloud real time as the drone's flying Who here is flying UAS right now with LIDAR? Any hands a few hands so the nice thing about the L1 and we're bringing it to the L2 as well Is that you're getting that live readout while the drone is flying so that you know when you go back to the office You should be able to process that data rather than finding out 24 hours later when you're back at the office Maybe I should have changed some flight parameters during capture. So you're getting all of that live feed in Also a big big aspect is the superior penetration So we're looking at a much much smaller beam size as well as putting out many more points per second We will be diving in on that as well We do have an improved detection range over the L1 now We're all the way up to 250 meters at 10% reflectivity and you can go much much higher than that if you have a more reflective surface Nice improvement over the L1 at 10% reflective Reflectivity being only 190 meters. So two weeks ago when I was flying in Guatemala Capturing some very very dense jungle DTMs. We were still able to fly 180 meters off the ground and still capture a lot amount of area and one scan and Be able to still hit the ground and then we also have an upgraded visual camera So the L1 the previous generation was using a very similar camera to the Phantom 4 Pro or the Phantom 4 RTK now we have an upgraded camera solution with the Mavic 3 Enterprises camera built into the L2 the biggest difference there is the capture speed The capture speed of the visual camera is now at 0.7 seconds Which means we never have to slow the drone down to make sure that the visual camera is keeping up with our light Capture now that camera is just clicking off every 0.7 seconds and we can fly the drone 15 meters per second and still get significant overlap Now I talked a lot on this slide. Let's dive deeper into each one of these aspects So let's start with high precision Here's a comparison of the L1 to the L2 and a number of different factors on why we're seeing improved accuracy Number one being that LiDAR module I'm not sure if I'm gonna be able to get a laser on a TV not quite So the LiDAR module you can see there from the Zenmue cell to the Zenmue cell one We're seeing a little bit of improvement going down to two centimeters accuracy But that's also capturing at a hundred and fifty meters instead of a hundred meters So once again that transitions into efficiency flying higher flying larger area while maintaining the same Accuracy as well as the same sampling distance Also, we have a new and improved INS inertial navigation system very similar to an IMU We don't have much difference on the The roll but the pitch and the yaw Actually have it backwards so the pitch and roll is actually the same But we do see significant improvements on the yaw That's one area of the L1 that we wanted to make sure to improve was that overall positional accuracy We were looking at 10 centimeters for horizontal at 50 meters with the L1 And now we're at down to five centimeters flying three times as high with that LiDAR sensor So we talked about being able to capture Did I skip a slide? Let me double check Nope, I think we're good cool So we talked about improved accuracy the next biggest thing is going to be that improved penetration With the L1 the maximum amount of points that we could output was 480,000 we could either do two returns at 240,000 points, which is pretty dense or three returns But you had to bump it down to 160,000 points now. We're getting 240,000 points or 240 kilohertz on all five Returns every second where you're capturing 1.25 million points, but we're also getting that depth We improved from the L1 to the L2 going from three returns to five returns And you'll show we'll show you why in a second How important that fourth and fifth return has been to getting a really good DTM in Superdense vegetation You also notice the beam divergence That's one point about accuracy too as we have a more accurate laser with a fifth the size beam divergence So now we when you're creating those curbs and gutter on the construction site Now we're going to see a lot less noise and a lot sharper curb when we're analyzing the data So what does that look like with more penetration in some of that deep vegetation? Well the L1 and the L2 side-by-side when we're really looking at those darker areas That's going to be where vegetation is overhanging and right now we're viewing just the orange Which is going to be just the ground so we're seeing a much more dense point cloud Even at the same capture rate so 378 points per square meter versus 362 But so many more of those points are actually hitting the ground Overall, so we're getting that we're getting a lot more discrepancy there getting all the way through the fourth and the fifth point So we talked about accuracy. We talked about better penetration We talked about a better laser that has a better beam divergence all of that correlates to higher levels of efficiency if we were to compare the L1 to the L2 and you were planning missions with the same Accuracy requirements as well as point cloud density now. We're able to fly with the L2 Five times as much data capture as the L1 so typically with the L1 We're maxing out at about point five Kilometer squared whereas you're getting that same density with the L2 as well at two point five square kilometers one other aspect I want to note about The efficient data collection one more raise of hands who here is specifically utilized the L1 before a Couple hands That system along with some others does require the system the payload to be on for at least five minutes to do an IMU Calibration we don't need to do that anymore So as soon as the drone is on as soon as it's starting to log RTK data or you're connected to a base station You can write that mission and it's ready to go and start capturing data. So even more efficiency is there Also, we have an enhanced user experience some of this rolls in from capturing from the L1 But we do have that live point cloud view now So on the controller you're actually seeing the points start to build You can then also click into an interactive 3d mode and start pinching and zooming all around a Live point cloud that continues to be built in as you're analyzing it So by the time you leave the field, you know that you should be able to process that data within Terra and get accurate results out We also are generating a field work quality report every time the drone comes home Whether it's a battery swap or the mission is complete It's also going to give you out this quality report and maybe something happened that you did not notice during flight You may have positional act inaccuracies where you can then go back and re queue the last mission if you did notice some issues And the field work quality report came back poor You can also go and view those 3d models that you've captured at a later time within the point cloud library One final aspect that I want to highlight Just a second is going to be the post-processing It's one thing To be able to create a very high-quality piece of hardware for LiDAR But we also need to be able to output an LAS file that you're going to be able to utilize in different feature or platforms Terra solid virtual surveyor topo dot whatever that platform may be we want to make it as streamlined as possible So with the L2 any purchase that you have on an L2 you get a lifetime Subscription to DJI Terra So you have a lifetime license that you're going to be able to run all of the stuff that you see today through and Then at the end you're going to be able to output output an LAS file After you've done all of this trip alignment and everything else So we just wanted to highlight you don't have an An additional cost every year that you got to keep utilizing this sensor You get all the software up front for free that you're going to need to use for the lifetime of the sensor Also a couple feature upgrades for Terra to here one of them is going to be I apologize I'm losing my voice from the show The accuracy optimization So if you toggle that on on point cloud optimization It is kind of like strip alignment But we have some extra AI features to help align and as you can notice from turning it off to turning it on Each pass that we're capturing lidar that strip alignment is much much smoother from one strip to the next Also, we do have the opportunity to do a smoothing point cloud function That's the next two lines on the very bottom there With it toggled off and toggled on so you may have noticed with the L1. There's a significant Well, I shouldn't say significant, but sometimes there's a bit of noise One aspect that we are able to do with Terra is help smooth out some of that noise But now that we have L2 capturing really really clean now we're able to create a very very precise Elevation profile you can notice that it's about a third the size of the elevation profile If you do turn the smoothing on So they can have a really crisp clean line Some of the other features that we did bring to DJI Terra If you are looking to generate the digital terrain model directly out of Terra and you don't want to use a third-party resource You can do that now as well. We're now classifying between vegetation and non-vegetation So if you just need to generate that DTM, you're perfect We also are able to generate DEMs if that's the main output you're looking for The next version of Terra that we're releasing over the next couple weeks Also, we'll have point cloud profile analysis So being able to click over a looks like a forestry area or a densely wooded area You then get that side profile and you can start measuring tree height You can start separating some of your trees out and you can understand elevation topography better And of course you find out You wrap up with a point cloud quality report overall That's going to give you just what was your processing quality from your data captured then process in the Terra What should what type of outputs you should be looking for? So we just talked a lot about the L2 and I just threw a lot of specifications at you But let's talk about how we've been actually able to get these sensors out in the field and Really start to adapt and use them We're gonna be talking about three different use cases. The very first one is power line digitization a very Important use case within lidar something that photogrammetry really cannot do or you're gonna require something with a point cloud Overview we have a customer in Indonesia that we are able to get an L2 beforehand for some testing We were able to fly this you can already see we were flew this twice as fast with the L2 And we're still able to get a cleaner point cloud and be able to reconstruct Because we have that fourth and fifth return be able to reconstruct some more of those lines that are attached to the main tower there So you're flying at 15 meters per second compared to seven meters per second We also had a gimbal angle of straight down And we did we did just fly on dual return So we didn't need a lot of vegetation poke and then a flight height of a hundred meters So still off the very top of the tower. We're looking at at least 60 to 55 meters off the top of the tower When we really start to analyze what that point cloud looks like you're really are able to Reconstruct that line much much cleaner on the left hand side. You can see the Right over here some of the those four lines and the eight lines that are connected the main tower When you compare that to the L1, we just really didn't get good reconstruction It's not that the lines are as blurry or have as much noise We did see that problem But also you're not getting a whole lot of reconstruction down below and you're really not getting a whole lot of reconstruction on the smaller aspects of the tower where you are with the L2 One final note you can kind of see how wide the swath was here We flew this just one pass if we didn't even have to necessarily come back to be able to generate this output So if you're looking at efficiencies being able to just fly and drive next to it or do hops You can improve efficiency as well So we move on from power lines Into construction a very low hanging fruit is what I typically say when you merge light our Construction often times when it comes to civil we're looking for what is that digital terrain model where we need a light our sensor Because we need to hit ground Overall what we found with a site in Korea They're able to filter out the ground points create a very very dense point cloud that you can see on the right hand side Or on the left hand side my right hand side We were able to generate contour lines from that DTM as well and then start analyzing the cut fill nothing too Revolutionary, but we did want to highlight the main improvements on accuracy from the L1 to the L2 So when we compare comparatively on cutting fills on fairly large stockpiles And fairly odd shaped stockpiles with very odd base planes. We still were seeing a lot less Margin of error compared to the L1 and Finally this one I'm going to spend a little bit more time on just because I was able to go out and help capture it This is a historical documentation in Guatemala happened two weeks ago It's some of the most dense jungle out in the world. I think on the next slide. We'll be able to see how dense it is What happened was and the reason why we were asked to come and do the mission was in the area There's a lot of ancient Mayan ruins and there's a lot of Causeways or their highways that lead throughout this very dense jungle But they've mapped out most the causeways But there aren't pyramids at every causeway that they've mapped out And so we still need to be able to find they probably led to a pyramid if there is a causeway there So let's go out there with a little bit more precision and try to find them they did Guatemala did go through and Contract out a manned aircraft with a very expensive LiDAR system on the belly of that manned aircraft Was able to do about two-thirds of the project And then the manned aircraft came down and they weren't able to recover the aircraft the LiDAR module or any of the data from that specific mission and so At the end of the day, they had about 120 square kilometers all in patches That needed to be patched in and so we were able to go into the dense jungle With four matrice 350s And l2s and then be able to start scanning that way. We were scanning at 160 Sorry 140 meters agl up to 180 meters agl We were flying at 12 meters per second We needed to penetrate that vegetation. So we set it to five returns Non-repetitive scan mode and we did set it at a pretty significant side lap to try to Be able to penetrate but overall What we found was we're in a very very remote area. There's no star link. There's no base station that you're getting Corrections off of you really need to either cut a hole in the jungle so that you can put your own base station there to Log ppk or cut a hole in the jungle or climb to the top of a pyramid to set up your rtk base station and fly from up there Overall, we were able to keep the drone in the air over 90 of the time and just keep doing battery swaps Even in the 90 percent or 90 degree heat fahrenheit And 80 humidity. We were able to keep that drone going Really be that workhorse. We were finding that each drone by the end of the day Was capturing around 15 square kilometers up to 20 square kilometers So depending on where you're located by the end of the two days we were able to get Uh, actually each day we were getting roughly around 30 square kilometers between the four drones that we were testing A lot of moving around But overall with the point clouds that we were able to capture Just to give a little bit of perspective the manned aircraft Of course it's capturing much larger areas than we are able to capture with the matrice 300 series Although four of them running at 180 meters at a time does capture quite a bit of data But they were hitting 40 to 50 points per square meter But only two to three points were hitting the ground So that dense that point cloud of the actual dtm Which is what we need to find the actual pyramids and the ruins is to remove all the trees We weren't able to do that very accurately the l1 Sure, we have more points than manned aircraft because we're flying lower and a bit slower But with triple return We still weren't able to quite hit the ground and we are only getting a few sample points But with the l2 overall we are sampling a lot higher Because we're getting out three times as many points per second But now that we have the fourth and fifth return We're getting 40 to 50 points hitting the ground And now the dtm comes out Pretty accurately and you're really we're getting a lot of points on the ruins to be able to build that dtm Those are the three different use cases For that we talked about at the covering the l2 At this point I ran a little bit quick. We have about five more minutes So if there are any questions happy to answer those some other participation at inner geo That I want to highlight Tomorrow we do have another speech on the main stage 10 a.m Talking about opportunities for effective data acquisition for the digital twin. We also have a partner bam tech Jared is going to be speaking tomorrow talking a bit more of their use of the l2 They did a 73 page white paper talking just nuts and bolts on the accuracy of the l2 Really running it through its courses. I would recommend that Also a shout out to terra solid They do a make sure to stop by their booth if you are using their platform already They are developing an in l2 integration guide on how to start using this data If you guys are able to go ahead and purchase it l2 I know the very first question is going to be price I'm not a salesperson But I will say and i'm also based on the north america team. So I don't know in euros But we did come out with the price at $13,599 I believe so overall The class point being a $13,000 usd Lidar module with imu with gimble. That's all integrated with the 350 It's really hard to beat that price point So we're excited to see how you guys might utilize the l2 feel free to stop by the dji booth I'm not good with booth numbers. It's that direction Uh, but overall we'd love to be able to talk more on the l2. Thank you all for joining