 So hello everybody Guten Morgan. This is Freda. I'm leading the solutions engineering team from DJI Enterprise Today, I'm delighted to be here to present our latest LiDAR payload DJI L2 and So I'm also co-presenting with our partner Jared from BEMTAC our partner in the United States So let me get started. Yeah Okay Can you hear me? Can you hear me? All right. I'm up on the microphone now. Great It is a pleasure to be here. Thank you Freda. Thank you DJI It is a pleasure for BEMTAC to be represented here in Energio We are a company in the United States But we had the privilege and the honor to go through and do the first validation of the L2 system Prior to release as you saw on Tuesday For those who are not aware DJI has been in this industry for 15 years with constant innovation And we saw that on Tuesday with the latest release of the DJI L2 Across the the last five six seven years We have seen innovation on the platform themselves the sensors the technology And finally with the latest laser system we get to see where we can go on small form factor LiDAR The latest is the Zenmuse L2 You saw the specs it is improved accuracy improved capability improved performance improved navigation Improvements across the board, but we needed to have a valid check of this data. It cannot just be taken at What is written that written down? We needed to go through a validation process through multiple testing environments With that we were able to build a testing structure that gave us something against industry standards It gave us something against multiple locations multiple environments and we tested it in every possible situation We could prior to release we continue to test it today We were flying out yesterday and we plan to keep flying for the next few months To prove the performance of this system In the United States, we are lucky enough to have an industry standard from ASPRS the American Society for photogrammetry and remote sensing most recently they updated their guidelines for mapping in 2023 back in August and we adapted those Specifications and those standards to our testing criteria for what we did with the L2 Through that we were able to do internal precision validation within swath swath the swath as well as an absolute Accuracy validation across multiple sites in the first site you can see a runway there We went out to Alameda a naval air station and we tested on a very flat Plainer surface to see how the quality and precision of the data was coming from the system Through that we used all of our testing and methodologies with Terra solid software Which has been doing point cloud intelligence for more than 30 years We were able to run multiple routines and use the software to Check every angle we possibly could on this data set In the very first test that we were able to do on this very flat surface We were able to assess using it within swath precision of less than one centimeter. That is repeatability across the plane this is Far just incredible compared to the L1. We were not expecting to see results of sub-centimeter But on this flat surface we were able to Validate that it was showing a planar surface incredibly accurate very similar to what you might see from high quality mobile systems One caveat was when we did get out to the site. We noticed that there was Some migrating geese and we did find a lot of debris on the runway, which gave us some interesting maximum Precision returns so we will have to do further testing on a more pristine test site Beyond that we were able to compare the within or swath to swath Analysis this is measuring the misalignment between one swath to the next in the same pass This gave us an idea of what the relative accuracy or what could be referred to as the relative accuracy of the system was In this we were able to see that we had an average offset visually of around one and a half centimeters from one line to the next But in the statistical results, we were at 1.15 centimeters mathematically from one plane to the next This was a great improvement over the L1 system in the past Finally we checked the absolute accuracy against ground survey with over a hundred and seventy check points We got a new mathematical calculation from the ASPRS for the vertical accuracy of the point cloud itself The vertical accuracy of the ground survey with a combined vertical accuracy of the first test of 1.19 centimeters absolute this is almost survey grade highest quality and Was beyond what we were expecting to see from this system In this category, we were able to use the Calculated results and identify the vertical accuracy class of the system Unfortunately because of those maximums that we saw from the migrating geese with we did have to go up one vertical accuracy class to 6.25 centimeters But if we were able to use the the average of the mean of the of the precision We could adjust this down to a 1.7 Vertical accuracy class which is near the top of the ASPRS standards To go beyond we wanted to test more sites and we wanted to do it with a survey that had a higher precision higher accuracy We we coordinated with a company of GBA engineering and survey firm in Kansas City They were able to go out and collect a validation site of survey to this It was less than 1 centimeter and had a misclosure of less than 1 centimeter I believe it was actually close to a less than 1 millimeter on the overall misclosure This gave us the highest vertical accuracy class of survey to allow us to do more validation tests with an absolute With the absolute best precision We flew more than 120 Different flights for this one site. We wanted to test it at different pulse reps We wanted to test it at different scan rates different scan patterns first returns last returns We wanted to run this system through the paces and give the best results and overall encompassing test we did give We did have a few problems that we would expect on any other flight because of weather or some other Environmental incident but other than that the tests went quite well in Throughout all of these different vertical tests all these different altitudes We did see a typical trend with the higher altitude the less accurate it gets But overall you would have an average accuracy within the swath of 1.36 centimeters. This is across the different flown altitudes up to 120 meters Beyond that we had a swath to swath test. This is where we saw a clear Trend between all of these with maximum and minimums the average swath to swath or RMSE DZ was 2.3 centimeters Again another incredible improvement that we were not expecting to see from the start After that we did have the final absolute accuracy tests This is where we saw a bit more of a spread when it came to the results But in general we had less than a centimeter up to below three centimeters of vertical accuracy across all tested altitudes and with an average Expected accuracy of one and a half centimeters from the system. This is Definitely a great improvement Beyond validating the accuracy and precision of the system we had to test the performance in the capabilities Where did in the beam divergence change increased precision? Where did we have additional returns and what could we expect for the results of these systems? How is it going to penetrate vegetation or the understory of vegetation? What are we going to expect when we fly the system every day and when everyone else flies the system every day and As was announced on Tuesday there was a great improvement of the beam divergence This allows for smaller more complex features to be discerned from others This allows you to do higher-grade engineering work and to go beyond what an entry-level LiDAR system is supposed to be able to do the improvement here was seen very clearly and Just based on the statistics and the actual published results the the DJI L2 does fall in line with some of the top Level ULS systems available on the market today There is only a few or a handful of systems that might be more improved than the DJI and they come in at a much higher cost One of the things that we were very proud to see was an improvement on the curb detection Through manual measurement. We validated that the DJI L2 was able to accurately Measure and to model the curbs that we were flying over this was similar to a mobile system and honestly at the Millimeter scale that we were seeing any kind of discrepancy that could be human error on measurement on our part or on the software's part It is minuscule so you have a very clear detection of Defined features in the vertical with this new system Beyond that we saw a greater contrast in intensities with more clarity This will allow you to do an additional feature extraction from the intensities alone and improve automation of any kind of performance with classification Feature extraction modeling that you would be doing in CAD environments Beyond that we were very impressed by the ray tracing of the RGB. This gave us a clear Evidence of where certain types of materials would differ inside of the vertical plane not just on nadir So you were no longer doing just a nearest neighbor Point cloud fusion you were actually to do ray tracing which gave you definition of the color on the point cloud Going into the the five returns with the vegetation penetration We were seeing clear and defined structure underneath vegetation canopy branches trees Into the understory which allowed us to do automatic classification of different trees grouping then segmenting through complete automation What we were able to see with the vegetation improvement or with those five returns was a clear definition of the structure underneath canopy drainage and any kind of Trail that we could imagine we were able to see the actual trails that we saw from animals walking through this heavily forested area with high leaf count Beyond that we did test it in crops where we were trying to test the The return separation or the the vertical resolution of the returns What we were able to find out was that you're looking at about point eight meters of return separation This will allow you to do higher and higher crops So high yield corn and you're still going to get returns at ground even with the smaller in crops like soy You're going to get penetration through the actual vegetation purely on the density alone of this system flying at the altitudes that you are in Conclusion we determined that the L2 was a vast improvement over the L1 through the improved performance the technical specifications the returns as well as the beam divergence We saw this system as going beyond entry level We saw this system going beyond what we were expecting and anyone else would probably expect that a system of this Price point and we were surprised to see that this system was going to be of an incredibly useful tool for surveying and mapping applications in the future this Validated to this point the accuracy that was being published for this system and it does validate The accuracy and precision of the system into the future We expect to do more tests and we expect to keep this moving forward But until then I thank you for your time and I do appreciate you showing up if you would like to get any more information on DJI or BAM tech you can also download the actual validation White paper and results as long as well as data if you go to this cue car code here. Thank you very much