I think next time you should start with a disclaimer that you are part owner of the company that makes the R2a. The data could be seen as skewed or biased when coming from someone who has such a conflict of interest. Good video otherwise.
First, I am not a lidar expert. I am a professional land surveyor, and I have been using the L1 for about a month now and have been testing on several different projects. I find your assessment perplexing. I find the “fuzziness” to be about 0.50 feet if you pick from max to min. If you select from high mean to low mean, the fuzziness drops to about 0.22 feet, a far cry from 1.8 feet. Next, I did not get any of the undulations that you are referring to in the video. Did you process in DJI Terra with the optimized point cloud accuracy turned on? I tested with and without, and the data will show some roll without processing with this option turned on. I did it mainly to compare results with post-processing in TerraMatch. I had a friend and lidar expert (certified photogrammetrist) whose assessments that Terra did as good a job as TerraMatch and is considerably easier to operate. My experience with the L1 with multiple projects is an overall RMSE of about 0.23 feet. According to ASPRS standards, this is plenty accurate for 1-foot contours, unlike your statement that the L1 was only suitable for 2-10 feet contours. Next, you fail to mention the total integration of the unit with the drone system. There is no turning on the LiDAR sensor, turning on the drone, manual calibration, clipping data post-flight. I found the total integration for the drone much preferable to other systems I have used. I have shared data with two photogrammetrist, and they both stated that the data is better than most lower-end LiDAR units. Is it a Riegl? No, but it is 1/10 of the cost. If you need 0.05 feet accuracy, get a Riegl. In my opinion, as a professional surveyor, the L1 will satisfy about 90% of the requirements of most raw land surveyors. I am not a DJI fanboy, but until American manufacturers catch up, I will be flying an M300/L1/P1 unit for most of my work.
Great comment. When I show the RMSe value derived from using the rock surveyor software - I report an RMSe of 0.206 ft. This is of the surface created by the A.I. engine. So there are 2 separate measurements happening here. The derived surface and the system accuracy. In the system, you can't just delete points and say - if i just choose these points and measure the precision then its better. You need to include all the data. So in your assessment this would be the 0.5ft measurement. Also, you can't use terramatch for a replacement of the DJI Terra software. The terra software is a tightly-coupled PPK software that calculates the trajectory using the differential GPS + INS setup. This is different than terramatch, which takes the results from DJI Terra and performs a strip alignment to correct and boresighting issues and small random walk noise in the heading. And there is a ton of manual calibration with the L1 - it is now recommended to do the calibration pattern every 100 seconds or 1,000 m - whichever comes first. I do agree the system is priced AMAZINGLY!! and that is freaking awesome, and I do love the L1 and I am a fan boy of DJI. They have innovated time and time again and it has benefitted us all over the years. But just like when I bought my M210 + RTK on the first batch that came out - there were a lot of "caveats" to make it work well. Simply put, the RTK system wasn't truly functional until the DRTK-2 came out.
@@IndianaDrones, I know what the RMSE values are coming from. If I use the values without manipulating the data, my RMSE is much tighter than the 0.22 I reported. I have experienced some values less than 0.05 feet. This, however, is not a true reading of the accuracy of the survey. If I am going to produce a real topographic survey, my check will be to the "filtered" point cloud-derived surface as this is what I am assuming liability. As to the ton of manual calibration, this is simply not true. The drone self-calibrates at the beginning and end of a mapping mission or a corridor mission. While you do have to pause a mission and hit the calibrate button if your distance exceeds 100 seconds. If you are flying your mission at 10m/s, this is 1000 meters or 3,280 feet. So far, I have not had to worry about this. In my conversations with DJI, this calibration will be automated in the next DJI pilot release. Also, in my opinion, the D-RTK-2 has never been functional. I either fly off of a VRS, the RobotDot base station, or I use PPK, which I have verified works with the L1 as well. While I am sure the rock is a good device, I have not seen or used any LiDAR system that is so well integrated and easy to use.
Mark, you literally contradicted yourself in your first paragraph. You say the 0.22 is not a true reading, then turn around and say, but if you do the same thing it’s a true reading. The 0.22 IS from the filtered point cloud surface. Therefore your first and second statement in that paragraph are the same thing. So either you do report your accuracy on that deliverable or you don’t. It can’t be, don’t if Indy does it and do if Mark does it.
Great assesment. He also made it seem "Rock" was the one to get since it was priced in the middle. But funny, he forgot to mention he owns the company lol
"@@IndianaDrones, I'm not sure I understand your response. You say your measurement is reported from the Rock "A.I." surface. Not knowing what this is, I am limited in my understanding of what the Rock "A.I." surface is computing. The RMsE I am reporting is the computed from independent field surveyed measurements compared to the point cloud processed in TerraScan and then exported. The surveyed points then are "dropped" onto the final surface. I have tested on three separate surfaces and foliage areas, and I am getting a 0.21 to 0.26 RMS on three separate projects. I flew all three projects using the Leica SmartNet VRS in RTK mode. On each project, another surveyor measured the GCPs and check shots independent of my flying. In one project, the GCPs and check shots were measures using a Trimble VRS. My point was if I just dropped the points onto a "raw" point cloud, my values were less. Not a valid check, in my opinion. This will be my last response. I find your analysis to be contradictory to my findings, as I stated in my first reply. I have not experienced the undulations; the fuzziness I have measured is about 0.50 max min and more on the order of 0.2 feet when picking the mean. I was cutting profiles using your setting of 0.50 feet width. This is before any processing. I have not experienced any of the undulations you reported. I find your assessment to be lacking and biased. You did not give any of the settings for any of your drone flights. Where they all RTK, PPK, or a mixture of both? What altitude were all the flights conducted? What was the setting of each sensor? How many returns for each sensor. What type of pattern scan? If you are going to do an evaluation, try to use some sort of scientific methodology. Best regards.
Big fan and love your videos. Didn't realize you are part owner of RockRobotic until now so need to take that into account when viewing the reviews but you do seem very honest about your results. I'm the RS Manager for a survey firm and just bought the L1. This video gives me a little concern but I also realize the L1 is literally just coming out and I'm sure DJI will have some tweaks coming up to their DJI Pilot and Terra software to help correct some of what is being seen here. Regardless, I think the bigger thing we all need to take away from this is that in my 15yrs of working with aerial LiDAR and Photogrammetry Reigl has been king and you need to reverse mortgage your home to get one.... but now we are seeing sub $50k sensors that nearly rival it... I see this time as a tipping point for affordable, accurate lidar for the survey/engineering world and am myself, super stoked.
Fantastic assessment of these sensors, Harrison. It was just what many of us non-lidar folks are looking for this sort of assessment before making some big investments. Well done. Keep 'em coming! You probably don't care about the trolls who may eventually start nitpicking, but it may be worthwhile to inform those who aren't aware, that you are affiliated with Rock Robotic/Rock Cloud and the R2A used in the video. A "Full disclosure" thing. It would just add to your already considerable authenticity.
Absolutely! Also to be noted that it was this channel that made me make the R2A and the cloud! I really love comparing this stuff and flying robots with lasers shooting from their bellies! 😍
Very curious to see if you can do some more long term testing on the L1 with that undulation effect in the data. Very informative and this will influence my purchase decision regarding these systems.
Here's a summary of the accuracy metrics: Hillside Guestimated "Fuzz Height" L1: 1.81 ft R2A: 0.281 ft MiniLUX: 0.18 ft Hillside GCP Vertical Accuracy RMSE L1: 0.206 ft R2A: 0.107 ft MiniLUX: 0.113 ft Hillside GCP Standard Deviation Elevation Delta L1: 0.172 ft R2A: 0.104 ft MiniLUX: 0.107 ft Hillside GCP Mean Delta L1: 0.205 ft R2A: 0.008 ft MiniLUX: -0.11 ft Airport Guestimated "Fuzz Height" L1: 0.485 ft R2A: 0.092 ft MiniLUX: 0.052 ft
draganfly is undoubtedly the best drone company! It is providing support in ukraine by supplying ukrainian people with medicines, insulin and much more. with this it is increasing its visibility and brand awareness. has an experienced team and unique technology, currently very underestimated based on its potential!
After further reviewing this, it was disappointing to know that this video was about "The $180k LiDAR unit is too much, L1 is too little money and it sucks and of course your unit in the middle of the two is the one to get". A disclaimer that you own Rock next time would be better, also L1 specifications are pretty clear on their website.
Dude! You crushed it! Your "interesting" results highlight that the MASSIVE price difference in LiDAR sensors needs to be addressed by the manufacturers (Riegl, YellowScan, Quanergy, etc.).
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions. community.rockrobotic.com/
It’s every 1,000 meters or 100 seconds, whichever happens first. Yea, this is new guidelines that DJI is putting out there but I flew many many missions with the L1 and never produced specially different results
Thank you for putting the video together. Is the measurement of the .48 ft or around 13cm of “undulation” the total undulation over the area described, or did you find the total undulation high/low greater than that? ie. 13cm of undulation high to low over the total cross section area? or is it just within one measure of high to low? I mean, is it within the standard deviation for the L1 specs ie. 5-10 cm (if flown at 50m) and if the undulations are within that range the result of rounding or within the range of undulations on runway and other high low readings from the other sensors. Also I noticed in the “fuzz” test on the L1 measurements was of the outliers… yet they were about 6.xx cm, seems within spec.. am I wrong? It seems the L1 is less accurate, yes, but is it less accurate notably outside the posted specs? Or could it be the result of measurements on outlier tall grass blades, dips in the ground under the grass, or in the case of the runway actually 13cm +- accuracy instead of the posted 5-10 cm in spec… What was the altitude of the flight? As noted by another commenter, and as a potential customer, I find it off putting not to hear or have in description, that the host of this video is partial owner of the Rock Robotics (Rock Cloud, R1, R2A) … knowing this places the burden of specific supporting evidence ie on the host, and the video could be helped with a few more details on the flights: ie. altitude , process of calibration, flight duration… this additional data could make sense of the inaccurate data with the L1… not saying that your findings are wrong… because if correct, it’s pretty important to know, especially if it shows the DJI specs are wrong 😱. Yet per the provided info in the video, the accuracy may still be within the range of the manufacturer specs (ie DJI)… I just can’t tell because the flight details and operations that could affect the accuracy (height…) were not discussed.
All systems are flown at 60m AGL (200ft). It wasn’t until after I flew that DJI changed their flight specs to include MAX 50m, and refly a calibration flight every 100 seconds with a max length of 1,000 meters. But, I flew another 10 or so missions and never got better than 15 cm. This was the best I could accomplish, and that is at one location, not including this time dependent. But, all the data is there for anyone to see. I know DJI is working on enhanced IMU feature that includes a photogrammetry technique to increase the accuracy - but this only works in scenarios where photogrammetry works well… so in that case, I would recommend photogrammetry
@@IndianaDrones Thank you for responding and answering this question. 60M is not considerably outside the 50M. Wow! 15 cm is considerable... and if it was also the case at 50M...? would be outside the posted specs. Yet even if its corrected with every 1000M recalibration maneuvers that is a real deterrent and would seem to me to minimize the total area of acquirable data per battery! ...Imagine it trying to do the figure 8, recalibration maneuvers while working on a steep slope or cliff scan... hmmm....The only reference in specs on the L1 I could find was on the DJI L1 website spec sheet, which referred at the bottom of the page * to the 1000m (the test of 5-10cm accuracy was flown on 1000m distance at 50M, with no mention of recalibration every 1000M) The calibration specs that you provided are new to me and something I only read in another comment on TH-cam somewhere. Could you provide a link or whereabouts one might find this new calibration spec. information. If this is the new specs from DJI L1, this, my friend, would be a win in Rock Robotics direction... I just can't seem to find that info on the web... is it only available to resellers?
Great video, would love to see an updated version with the DJI updates over the last year to see if it's made any improvements on the accuracy of the data collected.
Thanks! Check out this article I wrote on how the ROCK Cloud Processing Software can help improve L1 accuracy. www.rockrobotic.com/post/improving-accuracy-dji-zenmuse-l1-rock-cloud
Most likely the imu unit inside L1 is an inferior quality compare to that of the other two systems. Or Terra is not robust enough to get rid of the undulation issue. Whatever it is, we are skipping the current L1 until they sort it out..50m agl and every 1000s calibration just do not cut it for productive topographical survey works..
If you are looking for a high end Photogrammetry Payload, get the L1. The more homework I've done on this unit the more I notice it's just an overlap of Photogrammetry Point-Cloud Data and a low end LiDAR Scanner. Just buy the R2A. Much better option. Especially for the cost of Investment and your Return On Invevestment (ROI).
@@kentthomas1045 ROCK-Robotics R2A LiDAR Unit and software on the Rockcloud. Those are the best with the Emlid A2 RTK Unit. Tell them Global Hawk sent you and you can get a discount and some free Acreage upload funding with your purchase.
With all that great tech it seems crazy that you have to hand calculate a variance using mouse and calculator. Seems like a select box w stats is in order. It would be nice to see the overall accuracy (e.g. standard error of your GCP locations), rather than just point estimates. That being said, great video, great information and great enthousiasm!
Great comparison Harrison, it was a blast to be out there. Are you experimenting with different targets in the field? I just read about pyramid targets in the latest issue of LiDAR Magazine.
Good video Harrison! This was a difficult site to survey. The boundary was a few magnitudes more difficult for a few reasons. But these are the only boundary problems I enjoy because of the challenges.
Great video, it is genuinely amazing how you guys persevere using feet and inches these days though when looking at something so involved with accuracy and precision.
So you processed the data from the L1 with Rock Cloud, Is there some work around that gets you past having to pull the data with DJI Terra, or are you just eating the yearly subscription on that just to pull the data?
Thanks, Indy. Another useful video from which one can see L1 is good for monitoring and inspection rather than survey jobs. I am sure DJI will improve their L1 gizmo fast. We market GVI LiAIR V70 here in Ru and Kz. It uses the same LiVOX Avia LiDAR and IMU by Novatel which is a great contributor into good results. A big brother Riegl is well fit but is surely overpriced. Solutions get smaller and more affordable. Crushing a heavy and two BMWs worthy bird is a disaster. Keep on doing this, dude.
Great Video, I really like your Channel. one question, do you know what is the Livox lidar model inside the L1 you have tested? it seems that DJI has changed the lidar inside from the MID-70 to the AVIA (more precise) and I don't know when this change happened. if your system used the MID-70, it could be interested to test the new version of the L1 to see if the precision has improved....
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions. community.rockrobotic.com/
There is something off with this comparison. First the point density in the accuracy reports are definitely off. L1 says 0 points per sq feet, R2A 49 points per sq feet, then 4 points per sqft. I would also imagine that the R2A will have better processing accuracy through its propriatary software than the other. But anyways the meet and potatoes airstrip comparison was a bad experiment the relative precision between the LiDARs means nothing that they in some cases are lower or higher doesnt tell you which one went low and which one went high, the real comparison would have been against the GCPs which you didn't show. I mean I do believe for the price of the L1 it will not be as accurate as 30K+ ones, and the L1 is not 17k its 13k since it came out. Also I would've also enjoyed the setting at which the L1 and the other LiDARs captured their data.
Hey Eduardo, the systems were on the same settings - repeat mode and 3 returns. You can look at the data for yourself as well and you will see the GCP are spot on the riegl and rock. The only other logical explanation is the riegl and the rock and the emlid base rover pair, which was set up over a known point provided by the usgs and transmitting corrections over ntrip are all wrong. The only logical conclusion would be that the one device was least accurate. With the rock, riegl, and emlid you can see the PDOP as well as cycle slips and thusly verify the trajectory. You can also see the forward-reverse / combined separation from the tightly coupled solution and verify that yes - in fact the solution is very good. Now after going through this thorough analysis I can feel confident in my words on the video. There is in fact an inaccuracy in the L1 system.
@@IndianaDrones Oh I do understand the L1 is a less accurate system, but maybe not as accurate as it seems in this video, I still don't understand the accuracy reports of 0 points per sq feet. And that being that possible the L1 is equally as accurate but at a lower elevation than the other two with just maybe 10% or even 20% increase in time flying. Then people can say well the L1 is accurate but it needs to be flown lower increasing my flying time. Your conclusion is that it is not survey grade, which I think is incorrect by your own data, even possibly more incorrect if you would fly the system considering its capabilities. tens of thousands of dollars less for a 20 percent decrease in efficiency can be a great system for many smaller survey firms to start in the LiDAR industry.
Thanks for the comparison, it doesn't seem like L1 is ready for the big time yet. Have you done any comparisons on their ability to penetrate tree canopy and vegetation?
I have, the Avia sense that it is based off of works surprisingly well! I had the R2A in Missouri a couple weeks ago and flew a 200 acre site and got great results. It’s not the L1 but the same fundamental laser scanner. I could put together a quick video on that if it would be helpful!
I am interested in this technology as an Instrumentation and Monitoring engineer. At what point are crossing the line between monitoring (ground movement) and surveying? I wouldn't want to be surveying without a license. The obvious answer is to hire a licensed surveyor, but it's not that easy as we work all over the US.
There is nothing wrong with using the equipment for your use case. The only time you need the licensed surveyor is if you are submitting your measurements to a governing body. Think of it like a tape measure. Any one on the construction site can use one, but if you are using that tape measure to submit plans to the county then they want those measurements to be signed and sealed by a certified individual
@@IndianaDrones Do you mean "submitting measurements to a governing body" for purposes of legal boundaries or as-built locations of structures? We submit displacement values to owners (DOT, County, City) all the time, but our measurements tell you if something moved, not where it is in the real world coordinate system.
Can you include accuracy reports in the deliverables section of the Airport data sets, just like the hillside ones? These reports were very helpful for comparison. Great video, Thanks!
Hey Avinash! Thanks for your feedback. Feel free to bring your questions and requests over to our ROCK Robotic Community discussion board. Our support team is ready to answer your questions. community.rockrobotic.com/
Great video! Thank you for this valuable comparison!!! Would you mind sharing what scan mode you used to capture the L1 data? Looking at the point density sheet of the report it seems like the L1 used the “non-repetitive” scone mode while the other 2 used a “repetitive” or line scan mode. Could this impact the fuzz that was found in the L1 data? And have you done any comparisons of the 2 scan modes of the L1 system?
Great question! The same scan mode was used. The repetitive, the non-repetitive is much much worse, and that’s not just on the L1.. it’s any system using the avia sensor. There’s some very interesting technical reasons, but it’s in the way the Risley prism works
thanks for video. I read Zenmuse L1 detection range "450m @ 80% reflectivity, 0 klx ;190m @ 10% reflectivity, 100 klx" and i don't understand. Can you explaine for me :)
You saved me a good chunk of money for now, so thank you. I wonder if this issue could be solved in a future software update or if it is entirely derived from the hardware within the L1? This is outside of my technical knowledge, any insights?
hi, If your Matrice 300 is paired with D-RTK2, do you still need to use GCPs to achieve better accuracy? What's the difference between using M300+DRTK2 vs M300+GCPs?
i’ve been watching the latest docs on finding new maya locations in dense jungle - after the drone flys the area they are able to remove vegetation - are these able to do that?
Hello Do you think Lidar could be use to detect oil pollution at sea ? I mean you think lidar could make difference of density beetween sea-water density and diesl-oil density ? Thnak you Gérald
Harrison/community: What would one use the L1 system for if it, as a lidar system, can not give survey grade results? Is this more of an introductory kit to provide volumetric measurements and decent 1-2' contours, and maybe higher quality 3D modeling? Thanks
Hello. Very cool your videos. What do you recommend for data processing with photos or LIDAR? A set of software and hardware tools. Do you have a video on this topic?
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions. community.rockrobotic.com/
Great video, very informative. I do have some questions though, as I am seriously considering purchasing the LiVOX AVIA sensor in either the LiAir70, L1, or Rock R2A configuration, and am quite concerned about the undulating errors you identified with the DJI L1. Q: Did you do those comparisons before or after you conducted a boresight error correction to the data? In either event, have you come up with any explanation to the errors? Q: Have you seen these same errors with the LiAirV (or V70) package? Thanks, and keep making these great videos.
Hey Michael! Thanks for your feedback. Feel free to bring your questions over to our ROCK Robotic Community discussion board. Our support team is ready to answer your questions. community.rockrobotic.com/
Heat Mirage Just curious, could the precision of the runway data be varying due to a heat mirage above the surface of the runway. Lasers can be refracted just as light can, so that may be the issue you're running into. Check what time of day you measured with the L1. Did you run it last? The runway is one big thermal mass, so it gives off the most heat later in the day. I may be wrong, but it's worth experimenting with.
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions. community.rockrobotic.com/
Thanks for watching! Feel free to bring your questions over to our discussion board at community.rockrobotic.com - you can find answers and get help there!
Hi, I need help choosing enough cheap LiDAR drone for a charity work. I would like to use it in refugee camps, unfortunately they just place the tents/sheds without careful thoughts and when the rainy season comes some areas with sheds become flooded. I am interested in something not fancy, not interested in detailed high resolution thing just need enough to figure out the elevation and which low points to avoid placing tents. Further idea is to "connect" the low areas and channel the rainwater to a man made pond which can be used for harvest. Thanks.
One of those moments again overwhelmed by information and knowledge of others about drones and everything related. Just started as a licensed pilot but damn still so much to learn, and where the hell to start XD?
A few questions/comments 1) is the DJI L1 drone named this due to its GNSS only recording the L1 carrier phase? If so, this would explain the undulation. 2) regarding horizontal accuracy comparing the GCP's, does your cloud software have the option to enter lever arms (distance offsets between camera, lidar, GNSS, IMU) for non-R2 systems? 3) Did you try generating control reports in another software like TSCAN? The rock cloud seems user friendly, but concerned it could be biased towards R2 data sets depending on the derived algorithms used. Interesting video, thanks for the content.
1. I don't think so! Check out DJI.com for more info on that. 2. Those values are entered in the creation of the .LAS. file, which is done before uploading to the ROCK Cloud. 3. They're not biased and you can manually check them in the ROCK Cloud... or any other software. Thanks for watching!
Umm Interesting.. thanks.. I am not as up on LiDAR as I should be but can the returns be isolated and what happens IF you perform a regression analysis on the raw data? How does your data look then? The anomaly in the data you suspect is a time differential, could that be from a reflective surface throwing off the data.. maybe water on the surface on or near the runway?
He literally flew over the exact same runway in the same day. most likely it is the GNSS solution or else just he lidar sensor itself was drifting in precision.
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions. community.rockrobotic.com/
Can you share some thoughts and opinions on the P1 in a future video maybe? Maybe the same on the M300 vs other aerial survey drone solutions. How does DJI stack up compared to the competition? Are they the best value for performance, or are there better solutions on the market that you would recommend?
Love the videos, the company I work for just got the L1 sensor and I am currently getting some missions lined out for testing. Do you have any idea what might be causing the errors you are seeing? I only ask because I am new to all of this and don't the RA2 and the L1 use the same lidar sensor? Thanks and keep up the good work.
Hi, thanks for a great video. I suspect the undulation on the L1 comes from the RTK GNSS receiver's calculation of the elevation, assuming that these measurements were done in RTK mode and not post processed. That is the only time related variable in the processing of the cloud. You should try to measure using GPS only and lock out the other constellations from the RTK solution. I agree, the wide fuzz on the L1 is a deal breaker.
thinking same thing. i bet they did RTK. PPK is way better way to position DJI's GNSS. i bet this was the on board rtk engine. When I do PPK and use all constellations I still get 8mm or better positions it seems. But I do it manually and with python scripts etc. Not user friendly.
I’m interested in learning more about this field of work and maybe starting a business how did you get your hands on a 180,000$ piece of equipment did you buy it outright or can you get a lone from the bank and wright it as a business expense?
When you find most customers that are interested in having data analysis for surveying their vegetation. They are trying to get accurate and reliable data. From what you just showed though unless your doing something to where customers would make use of this info like a bridge and consistent water level rising data or cable points,Grand Canyon etc. it seems like the L1 is what makes the most sense wouldn’t it? Because all of the data was within 1 ft or so of the other two which one cost 200k and the other 40?
I think next time you should start with a disclaimer that you are part owner of the company that makes the R2a. The data could be seen as skewed or biased when coming from someone who has such a conflict of interest. Good video otherwise.
Agreed
Thank you for your feedback Jake!
Man..the quality you put on your videos is astonishing!!! Thank you for all the good work
Thank you!
First, I am not a lidar expert. I am a professional land surveyor, and I have been using the L1 for about a month now and have been testing on several different projects. I find your assessment perplexing. I find the “fuzziness” to be about 0.50 feet if you pick from max to min. If you select from high mean to low mean, the fuzziness drops to about 0.22 feet, a far cry from 1.8 feet. Next, I did not get any of the undulations that you are referring to in the video. Did you process in DJI Terra with the optimized point cloud accuracy turned on? I tested with and without, and the data will show some roll without processing with this option turned on. I did it mainly to compare results with post-processing in TerraMatch. I had a friend and lidar expert (certified photogrammetrist) whose assessments that Terra did as good a job as TerraMatch and is considerably easier to operate.
My experience with the L1 with multiple projects is an overall RMSE of about 0.23 feet. According to ASPRS standards, this is plenty accurate for 1-foot contours, unlike your statement that the L1 was only suitable for 2-10 feet contours. Next, you fail to mention the total integration of the unit with the drone system. There is no turning on the LiDAR sensor, turning on the drone, manual calibration, clipping data post-flight. I found the total integration for the drone much preferable to other systems I have used. I have shared data with two photogrammetrist, and they both stated that the data is better than most lower-end LiDAR units. Is it a Riegl? No, but it is 1/10 of the cost. If you need 0.05 feet accuracy, get a Riegl. In my opinion, as a professional surveyor, the L1
will satisfy about 90% of the requirements of most raw land surveyors. I am not a DJI fanboy, but until American manufacturers catch up, I will be flying an M300/L1/P1 unit for most of my work.
Great comment. When I show the RMSe value derived from using the rock surveyor software - I report an RMSe of 0.206 ft. This is of the surface created by the A.I. engine. So there are 2 separate measurements happening here. The derived surface and the system accuracy. In the system, you can't just delete points and say - if i just choose these points and measure the precision then its better. You need to include all the data. So in your assessment this would be the 0.5ft measurement.
Also, you can't use terramatch for a replacement of the DJI Terra software. The terra software is a tightly-coupled PPK software that calculates the trajectory using the differential GPS + INS setup. This is different than terramatch, which takes the results from DJI Terra and performs a strip alignment to correct and boresighting issues and small random walk noise in the heading.
And there is a ton of manual calibration with the L1 - it is now recommended to do the calibration pattern every 100 seconds or 1,000 m - whichever comes first.
I do agree the system is priced AMAZINGLY!! and that is freaking awesome, and I do love the L1 and I am a fan boy of DJI. They have innovated time and time again and it has benefitted us all over the years. But just like when I bought my M210 + RTK on the first batch that came out - there were a lot of "caveats" to make it work well. Simply put, the RTK system wasn't truly functional until the DRTK-2 came out.
@@IndianaDrones, I know what the RMSE values are coming from. If I use the values without manipulating the data, my RMSE is much tighter than the 0.22 I reported. I have experienced some values less than 0.05 feet. This, however, is not a true reading of the accuracy of the survey. If I am going to produce a real topographic survey, my check will be to the "filtered" point cloud-derived surface as this is what I am assuming liability.
As to the ton of manual calibration, this is simply not true. The drone self-calibrates at the beginning and end of a mapping mission or a corridor mission. While you do have to pause a mission and hit the calibrate button if your distance exceeds 100 seconds. If you are flying your mission at 10m/s, this is 1000 meters or 3,280 feet. So far, I have not had to worry about this. In my conversations with DJI, this calibration will be automated in the next DJI pilot release.
Also, in my opinion, the D-RTK-2 has never been functional. I either fly off of a VRS, the RobotDot base station, or I use PPK, which I have verified works with the L1 as well. While I am sure the rock is a good device, I have not seen or used any LiDAR system that is so well integrated and easy to use.
Mark, you literally contradicted yourself in your first paragraph. You say the 0.22 is not a true reading, then turn around and say, but if you do the same thing it’s a true reading. The 0.22 IS from the filtered point cloud surface. Therefore your first and second statement in that paragraph are the same thing. So either you do report your accuracy on that deliverable or you don’t. It can’t be, don’t if Indy does it and do if Mark does it.
Great assesment. He also made it seem "Rock" was the one to get since it was priced in the middle. But funny, he forgot to mention he owns the company lol
"@@IndianaDrones, I'm not sure I understand your response. You say your measurement is reported from the Rock "A.I." surface. Not knowing what this is, I am limited in my understanding of what the Rock "A.I." surface is computing. The RMsE I am reporting is the computed from independent field surveyed measurements compared to the point cloud processed in TerraScan and then exported. The surveyed points then are "dropped" onto the final surface. I have tested on three separate surfaces and foliage areas, and I am getting a 0.21 to 0.26 RMS on three separate projects. I flew all three projects using the Leica SmartNet VRS in RTK mode. On each project, another surveyor measured the GCPs and check shots independent of my flying. In one project, the GCPs and check shots were measures using a Trimble VRS. My point was if I just dropped the points onto a "raw" point cloud, my values were less. Not a valid check, in my opinion.
This will be my last response. I find your analysis to be contradictory to my findings, as I stated in my first reply. I have not experienced the undulations; the fuzziness I have measured is about 0.50 max min and more on the order of 0.2 feet when picking the mean. I was cutting profiles using your setting of 0.50 feet width. This is before any processing. I have not experienced any of the undulations you reported. I find your assessment to be lacking and biased. You did not give any of the settings for any of your drone flights. Where they all RTK, PPK, or a mixture of both? What altitude were all the flights conducted? What was the setting of each sensor? How many returns for each sensor. What type of pattern scan? If you are going to do an evaluation, try to use some sort of scientific methodology. Best regards.
Big fan and love your videos. Didn't realize you are part owner of RockRobotic until now so need to take that into account when viewing the reviews but you do seem very honest about your results. I'm the RS Manager for a survey firm and just bought the L1. This video gives me a little concern but I also realize the L1 is literally just coming out and I'm sure DJI will have some tweaks coming up to their DJI Pilot and Terra software to help correct some of what is being seen here. Regardless, I think the bigger thing we all need to take away from this is that in my 15yrs of working with aerial LiDAR and Photogrammetry Reigl has been king and you need to reverse mortgage your home to get one.... but now we are seeing sub $50k sensors that nearly rival it... I see this time as a tipping point for affordable, accurate lidar for the survey/engineering world and am myself, super stoked.
Have you seen these similar results or have their been improvements in processing or software in the last 7 months?
Also wondering if the undulations have been corrected.
How has the L1 been working out for you? Have you noticed the same level of inaccuracies mentioned in the video?
Fantastic assessment of these sensors, Harrison. It was just what many of us non-lidar folks are looking for this sort of assessment before making some big investments. Well done. Keep 'em coming! You probably don't care about the trolls who may eventually start nitpicking, but it may be worthwhile to inform those who aren't aware, that you are affiliated with Rock Robotic/Rock Cloud and the R2A used in the video. A "Full disclosure" thing. It would just add to your already considerable authenticity.
Absolutely! Also to be noted that it was this channel that made me make the R2A and the cloud! I really love comparing this stuff and flying robots with lasers shooting from their bellies! 😍
How very disappointing after watching this. Just ordered the L1 😞
@@thegooserider7245 Don't worry. This content will be useful to you. th-cam.com/video/NVcRdliaBcU/w-d-xo.html
@@IndianaDrones This is most... how do you say?... "interesting"...
Hi, the problem on the L1 module could be imu calibration problem in the roll.
Very curious to see if you can do some more long term testing on the L1 with that undulation effect in the data. Very informative and this will influence my purchase decision regarding these systems.
A lot has changed in the last couple years. Check out my new comparison video here: th-cam.com/video/dgQKaEZORro/w-d-xo.html
Thanks for taking the time to compare and share!
Here's a summary of the accuracy metrics:
Hillside Guestimated "Fuzz Height"
L1: 1.81 ft
R2A: 0.281 ft
MiniLUX: 0.18 ft
Hillside GCP Vertical Accuracy RMSE
L1: 0.206 ft
R2A: 0.107 ft
MiniLUX: 0.113 ft
Hillside GCP Standard Deviation Elevation Delta
L1: 0.172 ft
R2A: 0.104 ft
MiniLUX: 0.107 ft
Hillside GCP Mean Delta
L1: 0.205 ft
R2A: 0.008 ft
MiniLUX: -0.11 ft
Airport Guestimated "Fuzz Height"
L1: 0.485 ft
R2A: 0.092 ft
MiniLUX: 0.052 ft
draganfly is undoubtedly the best drone company!
It is providing support in ukraine by supplying ukrainian people with medicines, insulin and much more.
with this it is increasing its visibility and brand awareness.
has an experienced team and unique technology, currently very underestimated based on its potential!
100% agree $DPRO is the best!
After further reviewing this, it was disappointing to know that this video was about "The $180k LiDAR unit is too much, L1 is too little money and it sucks and of course your unit in the middle of the two is the one to get". A disclaimer that you own Rock next time would be better, also L1 specifications are pretty clear on their website.
Dude! You crushed it! Your "interesting" results highlight that the MASSIVE price difference in LiDAR sensors needs to be addressed by the manufacturers (Riegl, YellowScan, Quanergy, etc.).
Thank you!
How was the L1 and the other two units positioned? Was it RTK? Or did you PPK it later in the office?
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions.
community.rockrobotic.com/
I also heard from CanDrone, that the L1 requires a recalibration (figure 8) after every 1000m of flight. Is this maybe the issue?
It’s every 1,000 meters or 100 seconds, whichever happens first. Yea, this is new guidelines that DJI is putting out there but I flew many many missions with the L1 and never produced specially different results
Thanks dude. I will buy L2 after DJI improvements heheeh
Thank you for putting the video together. Is the measurement of the .48 ft or around 13cm of “undulation” the total undulation over the area described, or did you find the total undulation high/low greater than that? ie. 13cm of undulation high to low over the total cross section area? or is it just within one measure of high to low? I mean, is it within the standard deviation for the L1 specs ie. 5-10 cm (if flown at 50m) and if the undulations are within that range the result of rounding or within the range of undulations on runway and other high low readings from the other sensors.
Also I noticed in the “fuzz” test on the L1 measurements was of the outliers… yet they were about 6.xx cm, seems within spec.. am I wrong?
It seems the L1 is less accurate, yes, but is it less accurate notably outside the posted specs? Or could it be the result of measurements on outlier tall grass blades, dips in the ground under the grass, or in the case of the runway actually 13cm +- accuracy instead of the posted 5-10 cm in spec… What was the altitude of the flight?
As noted by another commenter, and as a potential customer, I find it off putting not to hear or have in description, that the host of this video is partial owner of the Rock Robotics (Rock Cloud, R1, R2A) … knowing this places the burden of specific supporting evidence ie on the host, and the video could be helped with a few more details on the flights: ie. altitude , process of calibration, flight duration… this additional data could make sense of the inaccurate data with the L1… not saying that your findings are wrong… because if correct, it’s pretty important to know, especially if it shows the DJI specs are wrong 😱. Yet per the provided info in the video, the accuracy may still be within the range of the manufacturer specs (ie DJI)… I just can’t tell because the flight details and operations that could affect the accuracy (height…) were not discussed.
All systems are flown at 60m AGL (200ft). It wasn’t until after I flew that DJI changed their flight specs to include MAX 50m, and refly a calibration flight every 100 seconds with a max length of 1,000 meters.
But, I flew another 10 or so missions and never got better than 15 cm. This was the best I could accomplish, and that is at one location, not including this time dependent.
But, all the data is there for anyone to see. I know DJI is working on enhanced IMU feature that includes a photogrammetry technique to increase the accuracy - but this only works in scenarios where photogrammetry works well… so in that case, I would recommend photogrammetry
As for the undulation, this was on top of the “fuzz” I measured this at an additional 15+- cm. This must be from the drift in the imu.
@@IndianaDrones Thank you for responding and answering this question. 60M is not considerably outside the 50M. Wow! 15 cm is considerable... and if it was also the case at 50M...? would be outside the posted specs. Yet even if its corrected with every 1000M recalibration maneuvers that is a real deterrent and would seem to me to minimize the total area of acquirable data per battery! ...Imagine it trying to do the figure 8, recalibration maneuvers while working on a steep slope or cliff scan... hmmm....The only reference in specs on the L1 I could find was on the DJI L1 website spec sheet, which referred at the bottom of the page * to the 1000m (the test of 5-10cm accuracy was flown on 1000m distance at 50M, with no mention of recalibration every 1000M) The calibration specs that you provided are new to me and something I only read in another comment on TH-cam somewhere. Could you provide a link or whereabouts one might find this new calibration spec. information. If this is the new specs from DJI L1, this, my friend, would be a win in Rock Robotics direction... I just can't seem to find that info on the web... is it only available to resellers?
@@IndianaDrones Have you looked to see if following the calibration flight and altitude recommendations fixed the undulation problem?
@@jimbonevideo6941 that's a really good question and a cool topic for a new video, go for it Harrison
Hi Harrison, great job. Thank you. It is good to have a technical opinion of this level. Go ahead and thanks for sharing.
Great video, would love to see an updated version with the DJI updates over the last year to see if it's made any improvements on the accuracy of the data collected.
Thanks! Check out this article I wrote on how the ROCK Cloud Processing Software can help improve L1 accuracy.
www.rockrobotic.com/post/improving-accuracy-dji-zenmuse-l1-rock-cloud
Love the quality and content of these videos!
Glad you like them!
One of your best videos. Great to see the differences.
Thanks!
Thanks for the awesome comparison.
Will you be able to compare the Microdrones system as well at some point in the future?
Would love to!
Most likely the imu unit inside L1 is an inferior quality compare to that of the other two systems. Or Terra is not robust enough to get rid of the undulation issue. Whatever it is, we are skipping the current L1 until they sort it out..50m agl and every 1000s calibration just do not cut it for productive topographical survey works..
If you are looking for a high end Photogrammetry Payload, get the L1. The more homework I've done on this unit the more I notice it's just an overlap of Photogrammetry Point-Cloud Data and a low end LiDAR Scanner. Just buy the R2A. Much better option. Especially for the cost of Investment and your Return On Invevestment (ROI).
What's best for DEM creation?
@@kentthomas1045 ROCK-Robotics R2A LiDAR Unit and software on the Rockcloud. Those are the best with the Emlid A2 RTK Unit. Tell them Global Hawk sent you and you can get a discount and some free Acreage upload funding with your purchase.
@@globalhawkllc what would the final price be?
About Time. Great work.
Thank you for comparing this for us , Surveyors!
Anytime!
Great comparison, Harrison! Thanks for sharing!
Great comparison. A comparison between the Rock R2A and a Greenvalley international LiAir can also be very interesting.
Was a basestation use for RTK or PPK or was this flown as is?
Rock units, and Reigl use PPK. The L1 uses RTK
@@MDvkAir perhaps the vertical component (always the poorest) of the RTK solution contributes to these vertical undulations?
With all that great tech it seems crazy that you have to hand calculate a variance using mouse and calculator. Seems like a select box w stats is in order. It would be nice to see the overall accuracy (e.g. standard error of your GCP locations), rather than just point estimates. That being said, great video, great information and great enthousiasm!
Wow, it's quite compelling how closely the R2A matches with the MiniVUX, especially considering the vast price difference for these 2 systems.
Thanks for the information.
I think it’s probably coming for the INS that is not well managed by DJI.
It could explain long term oscillations.
Great comparison Harrison, it was a blast to be out there. Are you experimenting with different targets in the field? I just read about pyramid targets in the latest issue of LiDAR Magazine.
Thank you!
Good video Harrison! This was a difficult site to survey. The boundary was a few magnitudes more difficult for a few reasons. But these are the only boundary problems I enjoy because of the challenges.
Great video, it is genuinely amazing how you guys persevere using feet and inches these days though when looking at something so involved with accuracy and precision.
Thanks for watching!
So you processed the data from the L1 with Rock Cloud, Is there some work around that gets you past having to pull the data with DJI Terra, or are you just eating the yearly subscription on that just to pull the data?
Thanks, Indy. Another useful video from which one can see L1 is good for monitoring and inspection rather than survey jobs. I am sure DJI will improve their L1 gizmo fast.
We market GVI LiAIR V70 here in Ru and Kz. It uses the same LiVOX Avia LiDAR and IMU by Novatel which is a great contributor into good results.
A big brother Riegl is well fit but is surely overpriced. Solutions get smaller and more affordable. Crushing a heavy and two BMWs worthy bird is a disaster.
Keep on doing this, dude.
Thank you for your feedback!
Great Video, I really like your Channel. one question, do you know what is the Livox lidar model inside the L1 you have tested? it seems that DJI has changed the lidar inside from the MID-70 to the AVIA (more precise) and I don't know when this change happened. if your system used the MID-70, it could be interested to test the new version of the L1 to see if the precision has improved....
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions.
community.rockrobotic.com/
you used base rtk for correction?
RTK is a must for the L1 but all three were processed in PPK. L1 with Terra, R2A with PCMaster, and miniVUX with Inertial Explorer
@@IndianaDrones Could you process all of them with IE so no any difference from PP algorythm. Just curious how much difference could come.
There is something off with this comparison. First the point density in the accuracy reports are definitely off. L1 says 0 points per sq feet, R2A 49 points per sq feet, then 4 points per sqft. I would also imagine that the R2A will have better processing accuracy through its propriatary software than the other. But anyways the meet and potatoes airstrip comparison was a bad experiment the relative precision between the LiDARs means nothing that they in some cases are lower or higher doesnt tell you which one went low and which one went high, the real comparison would have been against the GCPs which you didn't show. I mean I do believe for the price of the L1 it will not be as accurate as 30K+ ones, and the L1 is not 17k its 13k since it came out. Also I would've also enjoyed the setting at which the L1 and the other LiDARs captured their data.
Hey Eduardo, the systems were on the same settings - repeat mode and 3 returns. You can look at the data for yourself as well and you will see the GCP are spot on the riegl and rock. The only other logical explanation is the riegl and the rock and the emlid base rover pair, which was set up over a known point provided by the usgs and transmitting corrections over ntrip are all wrong. The only logical conclusion would be that the one device was least accurate. With the rock, riegl, and emlid you can see the PDOP as well as cycle slips and thusly verify the trajectory. You can also see the forward-reverse / combined separation from the tightly coupled solution and verify that yes - in fact the solution is very good. Now after going through this thorough analysis I can feel confident in my words on the video. There is in fact an inaccuracy in the L1 system.
@@IndianaDrones Oh I do understand the L1 is a less accurate system, but maybe not as accurate as it seems in this video, I still don't understand the accuracy reports of 0 points per sq feet. And that being that possible the L1 is equally as accurate but at a lower elevation than the other two with just maybe 10% or even 20% increase in time flying. Then people can say well the L1 is accurate but it needs to be flown lower increasing my flying time. Your conclusion is that it is not survey grade, which I think is incorrect by your own data, even possibly more incorrect if you would fly the system considering its capabilities. tens of thousands of dollars less for a 20 percent decrease in efficiency can be a great system for many smaller survey firms to start in the LiDAR industry.
Fantastic assessment of these sensors!
Thanks for the information.
Please make comparisom between Livox and Reigl
there was a good one months before
Thanks for the comparison, it doesn't seem like L1 is ready for the big time yet. Have you done any comparisons on their ability to penetrate tree canopy and vegetation?
I have, the Avia sense that it is based off of works surprisingly well! I had the R2A in Missouri a couple weeks ago and flew a 200 acre site and got great results. It’s not the L1 but the same fundamental laser scanner. I could put together a quick video on that if it would be helpful!
@@IndianaDrones please do
My question is - how do you justify the 180 grand Lidar when comparing the results to the Rock's 40 grand sensor?
Personal preference I suppose! Check out my channel, there's been a lot of new advancements in ROCK LiDAR.
I am interested in this technology as an Instrumentation and Monitoring engineer. At what point are crossing the line between monitoring (ground movement) and surveying? I wouldn't want to be surveying without a license. The obvious answer is to hire a licensed surveyor, but it's not that easy as we work all over the US.
There is nothing wrong with using the equipment for your use case. The only time you need the licensed surveyor is if you are submitting your measurements to a governing body.
Think of it like a tape measure. Any one on the construction site can use one, but if you are using that tape measure to submit plans to the county then they want those measurements to be signed and sealed by a certified individual
@@IndianaDrones Do you mean "submitting measurements to a governing body" for purposes of legal boundaries or as-built locations of structures? We submit displacement values to owners (DOT, County, City) all the time, but our measurements tell you if something moved, not where it is in the real world coordinate system.
Recommend if you can generate surfaces and make your comparisons by generating a profile..using point cloud is not really ideal
Would PPK with GCP's throughout your mission site help prevent these undulations?
Excellent video mate.
the processed areas aren't the same??
97ac,96ac,70ac??
Can you include accuracy reports in the deliverables section of the Airport data sets, just like the hillside ones? These reports were very helpful for comparison. Great video, Thanks!
Hey Avinash! Thanks for your feedback. Feel free to bring your questions and requests over to our ROCK Robotic Community discussion board. Our support team is ready to answer your questions.
community.rockrobotic.com/
Thanks for the excellent information 👍
Glad it was helpful!
What is the file format of the 3D point space dataset? JSON?
Good work you put in this video!
Great video! Thank you for this valuable comparison!!! Would you mind sharing what scan mode you used to capture the L1 data? Looking at the point density sheet of the report it seems like the L1 used the “non-repetitive” scone mode while the other 2 used a “repetitive” or line scan mode. Could this impact the fuzz that was found in the L1 data? And have you done any comparisons of the 2 scan modes of the L1 system?
Great question! The same scan mode was used. The repetitive, the non-repetitive is much much worse, and that’s not just on the L1.. it’s any system using the avia sensor. There’s some very interesting technical reasons, but it’s in the way the Risley prism works
I would love to see this comparison done with the new BLK Drone.
Me too! Anyone have one ☝️ Let’s do this
@@IndianaDrones I know a guy…
It will be helpful when purchasing L1.
What do you think is necessary to get the correct height?
thanks for video. I read Zenmuse L1 detection range "450m @ 80% reflectivity, 0 klx ;190m @ 10% reflectivity, 100 klx" and i don't understand. Can you explaine for me :)
You saved me a good chunk of money for now, so thank you. I wonder if this issue could be solved in a future software update or if it is entirely derived from the hardware within the L1? This is outside of my technical knowledge, any insights?
Shouldn't you be able to apply a software correction for the RGB view to match it up with the LiDAR view?
Yeah! Now you can!
Is this available in India?
hi,
If your Matrice 300 is paired with D-RTK2, do you still need to use GCPs to achieve better accuracy?
What's the difference between using M300+DRTK2 vs M300+GCPs?
Hello from the DEMs, I got the contour lines and compared the 3 options. Strange result!
You mean that from the L1? Or in general? What are you seeing?
@@jrbcast123 I compared the contour lines of the three sensors
You are nothing but a LEGEND !!
Whoa!
i’ve been watching the latest docs on finding new maya locations in dense jungle - after the drone flys the area they are able to remove vegetation - are these able to do that?
Yes indeed! www.rockrobotic.com
Hello
Do you think Lidar could be use to detect oil pollution at sea ?
I mean you think lidar could make difference of density beetween sea-water density and diesl-oil density ?
Thnak you
Gérald
Typically the best sensor for this type of work is high-quality video and a trained eye.
How did you get a L1?
How big is that coffee cup..
Hi. May i Ask if RockR2A is available in the market? And where i can purchase it?
Hello Marco! Yes it sure is. You can learn more at www.rockrobotic.com/r2a or write to hello@rockrobotic.com and we'll get you more info!
What does the DJI say for the precision of L1?
Hey there - check out my new video for more information on the L1!
Harrison/community: What would one use the L1 system for if it, as a lidar system, can not give survey grade results? Is this more of an introductory kit to provide volumetric measurements and decent 1-2' contours, and maybe higher quality 3D modeling? Thanks
Exactly, it is used as an introductory kit.
@Indiana Drones Hey... introductory kits can still make 10's of thousands in the right hands!
did it get better over time? did you repeat the comparison maybe?
Technology always does! New comparison videos coming soon.
Brilliant video. Thank you very much!
Do you have subtitles in Portuguese?
Hello. Very cool your videos. What do you recommend for data processing with photos or LIDAR? A set of software and hardware tools. Do you have a video on this topic?
The ROCK Cloud! Visit www.rockrobotic.com/rockcloud for more information.
@@IndianaDrones Thank you.
What computer you use for processing the data?
(Processor, ram, etc)
Fully loaded Macbook Pro M1
About L1 "waves" - it could be "optimize point cloud accuracy" option, try to turn it off.
Do you think the final product of the L1 will be more accurate then the demo sample ?
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions.
community.rockrobotic.com/
What was your Ground sampling distance? for each drone system?
Learn more here: learn.rockrobotic.com/gcp-best-practices
thanks you for providing datapoints, to the point vid
You bet! Visit www.rockrobotic.com to learn more about my LiDAR company!
Great video, very informative. I do have some questions though, as I am seriously considering purchasing the LiVOX AVIA sensor in either the LiAir70, L1, or Rock R2A configuration, and am quite concerned about the undulating errors you identified with the DJI L1.
Q: Did you do those comparisons before or after you conducted a boresight error correction to the data? In either event, have you come up with any explanation to the errors?
Q: Have you seen these same errors with the LiAirV (or V70) package?
Thanks, and keep making these great videos.
Hey Michael! Thanks for your feedback. Feel free to bring your questions over to our ROCK Robotic Community discussion board. Our support team is ready to answer your questions.
community.rockrobotic.com/
@Michael Laura did you ever get an answer to this question?
Heat Mirage
Just curious, could the precision of the runway data be varying due to a heat mirage above the surface of the runway. Lasers can be refracted just as light can, so that may be the issue you're running into.
Check what time of day you measured with the L1. Did you run it last? The runway is one big thermal mass, so it gives off the most heat later in the day.
I may be wrong, but it's worth experimenting with.
By the way, I enjoy your videos. Keep it up!
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions.
community.rockrobotic.com/
THANKS , great job !
Is there a RIDAR that creates a map that can be transfer to f softwares, for example AUTOCAD?
Thanks for watching! Feel free to bring your questions over to our discussion board at community.rockrobotic.com - you can find answers and get help there!
Hi I’m just inquisitive but a interested newbie non scientist but I have a question. Does heat haze impact on the fuz and skew the contouring?
No! Learn more in my new article about L1 accuracy.
www.rockrobotic.com/post/improving-accuracy-dji-zenmuse-l1-rock-cloud
Thank you for a great video!
Maybe comparison of the als and tls would also be good.
Absolutely!
Hi, I need help choosing enough cheap LiDAR drone for a charity work. I would like to use it in refugee camps, unfortunately they just place the tents/sheds without careful thoughts and when the rainy season comes some areas with sheds become flooded. I am interested in something not fancy, not interested in detailed high resolution thing just need enough to figure out the elevation and which low points to avoid placing tents. Further idea is to "connect" the low areas and channel the rainwater to a man made pond which can be used for harvest. Thanks.
Hey there! Thanks for your interest. Please reach out to hello@rockrobotic.com - a member from my team will help you out!
One of those moments again overwhelmed by information and knowledge of others about drones and everything related.
Just started as a licensed pilot but damn still so much to learn, and where the hell to start XD?
Hello! Please feel free to bring any of your questions to our discussion board at community.rockrobotic.com - you can find answers and get help there.
INDIANA, perfect your case study... Do you think these ripples could be from DJI's RTK ?? since others use ppk...
I'm a bit disappointed the "shotgun" didn't make it into this.
All I could focus on was the piloting skills of whomever was flying that M210.
@@texasdronecompany9782 and that m600.....:-)
Now now, I’m sure that M210 ended up safe and sound
@@IndianaDrones ....on the ground..
A few questions/comments
1) is the DJI L1 drone named this due to its GNSS only recording the L1 carrier phase? If so, this would explain the undulation.
2) regarding horizontal accuracy comparing the GCP's, does your cloud software have the option to enter lever arms (distance offsets between camera, lidar, GNSS, IMU) for non-R2 systems?
3) Did you try generating control reports in another software like TSCAN? The rock cloud seems user friendly, but concerned it could be biased towards R2 data sets depending on the derived algorithms used.
Interesting video, thanks for the content.
1. I don't think so! Check out DJI.com for more info on that.
2. Those values are entered in the creation of the .LAS. file, which is done before uploading to the ROCK Cloud.
3. They're not biased and you can manually check them in the ROCK Cloud... or any other software.
Thanks for watching!
Umm Interesting.. thanks.. I am not as up on LiDAR as I should be but can the returns be isolated and what happens IF you perform a regression analysis on the raw data? How does your data look then? The anomaly in the data you suspect is a time differential, could that be from a reflective surface throwing off the data.. maybe water on the surface on or near the runway?
He literally flew over the exact same runway in the same day. most likely it is the GNSS solution or else just he lidar sensor itself was drifting in precision.
Thanks for your feedback! Feel free to bring your questions and requests over to the ROCK Robotic Community discussion board. Our support team is ready to answer all of your questions.
community.rockrobotic.com/
Zenmuse L1 just comes to the Lidar survey, it may take some time to make more improvements.
Can you share some thoughts and opinions on the P1 in a future video maybe? Maybe the same on the M300 vs other aerial survey drone solutions. How does DJI stack up compared to the competition? Are they the best value for performance, or are there better solutions on the market that you would recommend?
liair v70
Love the videos, the company I work for just got the L1 sensor and I am currently getting some missions lined out for testing. Do you have any idea what might be causing the errors you are seeing? I only ask because I am new to all of this and don't the RA2 and the L1 use the same lidar sensor? Thanks and keep up the good work.
Hey Justin! You should definitely reach out to us. We help with the L1 sensor and workflows and are here to get the best out of any system!
Excellent content!
Hi, thanks for a great video. I suspect the undulation on the L1 comes from the RTK GNSS receiver's calculation of the elevation, assuming that these measurements were done in RTK mode and not post processed. That is the only time related variable in the processing of the cloud. You should try to measure using GPS only and lock out the other constellations from the RTK solution. I agree, the wide fuzz on the L1 is a deal breaker.
thinking same thing. i bet they did RTK. PPK is way better way to position DJI's GNSS. i bet this was the on board rtk engine. When I do PPK and use all constellations I still get 8mm or better positions it seems. But I do it manually and with python scripts etc. Not user friendly.
I’m interested in learning more about this field of work and maybe starting a business how did you get your hands on a 180,000$ piece of equipment did you buy it outright or can you get a lone from the bank and wright it as a business expense?
Business loans for sure! Check out www.rockrobotic.com and write to sales@rockrobotic.com for more information!
Wonderful work, thanks my friend
Anytime!
When you find most customers that are interested in having data analysis for surveying their vegetation. They are trying to get accurate and reliable data. From what you just showed though unless your doing something to where customers would make use of this info like a bridge and consistent water level rising data or cable points,Grand Canyon etc. it seems like the L1 is what makes the most sense wouldn’t it? Because all of the data was within 1 ft or so of the other two which one cost 200k and the other 40?
First of all, no! Secondly, new price-point on the R2A ($19,995)!
Lidar Drones use this time 2022-2023 to the future Drone Detector
COMO ESTAS AMIGO UNA CONSULTA EL L1 ES BUENA EN LA PENETRACION DE VEGETACION
I would recommend the ROCK R2A LiDAR for scanning dense vegetation. Visit rockrobotic.com for more information.
What is the exact riegl model? On their web there a a few minivux. Also, for the L1, what altitude was the flight performed at? Thanks!