ACCURACY TEST with the LIXEL L2 Pro from XGRIDS at the HTW Dresden - Interview with Eric and Jannis

Summary

Highlights

• Eric and Jiannis discuss the L2 Pro scanner's outstanding accuracy at HTW Dresden. 🤓
• The scanner offers a wide range from 0.5 to 300 meters and up to 640,000 points. 🔍
• Tests revealed that ground control points improve scan accuracy by 4mm. 📏
• Comparison with other scanners showed the L2 Pro's strong performance and minimal deviation. 🥇
• Potential developments involve GPS integration for improved accuracy in future applications. 🌟

Key Takeaways

• The Lixel L2 Pro scanner boasts a remarkable accuracy range, making it suitable for various applications. 🎯
• Ground control points significantly enhance the scanner's precision, especially in topographical surveys. 🌍
• Cloud Compare and PointCap software played crucial roles in evaluating the scanner's performance. 💻
• The Lixel L2 Pro stood its ground against terrestrial scanners, showcasing minimal deviations in accuracy. 🚀
• Future possibilities include utilizing integrated GPS for enhanced surveying capabilities. 📡

Overview

Chapters

• 00:00 - 00:30: Introduction and Overview In the introduction, Eric from Laser Scanning TV and Jiannis introduce the topic of their discussion, which revolves around an accuracy test. They compare the larger L2 Pro version from Xcripts to the smaller K1 model previously discussed, indicating impressive results from the L2 Pro accuracy test. Viewers are encouraged to stay tuned for more details.
• 00:30 - 01:00: Interview with Jiannis In this chapter titled 'Interview with Jiannis', Jiannis introduces himself as a student studying geomatics at the State University of Applied Science in Germany. He mentions that he is currently in his sixth semester and has conducted accuracy testing on a device named A2 Pro, which is a scanner.
• 01:00 - 02:00: L2 Pro Specifications The 'L2 Pro Specifications' chapter discusses the capabilities of the L2 Pro scanner, focusing on its range, scanning frequency, and camera resolution. Depending on the version, the scanner can operate within a range of 0.5 to 300 meters. The 16-channel version of the scanner, specifically, supports a scan frequency of 320,000 to 640,000 points. Additionally, the chapter notes the camera's resolution is approximately 48 megapixels.
• 02:00 - 03:00: Software Overview The chapter gives an overview of a system called Lixel L2 Pro, a 16-channel SI system with a range of 120m. It's a smaller but faster version of a large scanner, capable of processing 320,000 points per second. The chapter also briefly mentions two software products from X grits, highlighting 'Lexus Studio', which is used for processing.
• 03:00 - 04:00: Indoor Accuracy Testing The chapter titled 'Indoor Accuracy Testing' involves the use of point clouds and Liix Cyber Color, which is a software that creates photorealistic 3D models based on 3D region splitting. The transcript discusses a video displaying a test field for indoor accuracy testing, highlighting the capability to switch between various data representations such as point clouds.
• 04:00 - 05:00: Ground Control Points Comparison The chapter titled 'Ground Control Points Comparison' discusses the advantages of using the X grid model, particularly its renowned caution splitting implementation. It highlights that the system is well-recognized not just for its optical or virtual features, but for the actual measurement system it provides. The chapter begins by exploring the measurement options available within the system, outlining the user's role in assessing these functionalities.
• 05:00 - 06:00: Point Cloud Software Analysis Summary of the 'Point Cloud Software Analysis' chapter: The chapter discusses the procedure for assessing the geometric accuracy of the L2 Pro software. It involves conducting accuracy tests in both indoor and outdoor settings. The indoor test field is described as a hall in the HTV driftton. The primary objective of the first task was to evaluate changes in room distances after processing, testing both scenarios with and without the use of grow control points. The results of these tests help determine the accuracy of the software in different environments.
• 06:00 - 07:00: Outdoor Accuracy Testing The chapter titled 'Outdoor Accuracy Testing' details a comparison study involving point clouds from the L2 and a terrestrial scanner called the fiber focus. The accuracy verification was conducted in automatic settings. The testing took place in a hall used for calibrating laser scanners, equipped with numerous control point markers which have highly precise coordinates.
• 07:00 - 08:00: Results and Observations The chapter discusses control points used in scans and their local coordinate systems, specifically highlighting four green points in a facility. This facility, described as very unstable, is a large concrete structure originally developed to house locomotives. However, the original purpose shifted over time as the university no longer possessed the necessary equipment, altering the facility's intended use.
• 08:00 - 09:00: Point Cap Software Analysis The chapter titled 'Point Cap Software Analysis' discusses the application of markers by mechanical engineers within a particularly stable building to test its structural integrity against vibrations. Moreover, the climate control within this structure is noted to be excellent. The core aim of the study was to assess whether the geometry of point clouds can be enhanced through the utilization of ground control points, and includes a comparative analysis to support this investigation.
• 09:00 - 10:00: Outdoor Reflective Target Testing The chapter discusses the process of outdoor reflective target testing using the terrestrial point cloud of the fiber focus for verifying accuracy outdoors. The methodology includes data acquisition with the L2 Pro device, processing the data using Lexus Studio both with and without crowd control points, and evaluating the data. The chapter highlights a specific focus on the comparison of distances within the collected data.
• 10:00 - 11:00: Conclusion and Future Prospects The chapter focuses on discussing the control points as shown in the upper right corner and their spatial locations using coordinates from Favo. The author conducts an analysis of the spatial distances between these control points and describes the methodology used for measurements in point cap, which involves creating sectional views and determining the middle points for accurate measurement. The chapter likely concludes with the outcomes of these analyses and provides insights or future directions based on the results gained from this process.
• 11:00 - 12:00: Closing Remarks The chapter discusses the calculation of distances using control points, noting the use of 15 different points with distances ranging from 2.8 meters to 40 meters. It highlights the measurement of deviation values from target distances in a point cloud, specifically without the influence of ground control points. The details are demonstrated through a chart, particularly focusing on the middle column's data.

ACCURACY TEST with the LIXEL L2 Pro from XGRIDS at the HTW Dresden - Interview with Eric and Jannis Transcription

• 00:00 - 00:30 Hello, Eric here from Laser Scanning TV company Laser Scanning Europe and I'm again here with Jiannis and uh Jiannis uh what we speak about this episode. Well, hi, I'm Jiannis and we did a accuracy test. Last time in the video we had the little one, the K1 and today we are speaking about the accuracy of the L2 Pro, the bigger version from Xcripts. I can say this uh we got really amazing results out and if you are not like to miss this then stay
• 00:30 - 01:00 tuned before we uh really uh make a sharp start here maybe you tell something about your person and and what what you done well hey I'm Jiannis and I'm studying geometric at the state university of applied science at in Germany I'm now in the sixth semester And yeah, I did some accuracy testing about the A2 Pro, which we can see here on the right side. Um, the scanner has a
• 01:00 - 01:30 range about uh 0.5 to 300 meter. Um, it depends on the version of the scanner. I used the 16 channel scanner and its scan frequency has 3u 20,000 points to 640,000 points and yeah we got a camera resolution about uh 48 uh MP and yes yeah I I think maybe I
• 01:30 - 02:00 precise this as we use the 16 channel version of Lixel L2 Pro is definitely an SI system and is having then the range from 120 m. This is the um this is the smaller version from the big scanner and have the speed from 320,000 points per second. So now we can go forward. Let's come to the two softwares from X grits. On the left side here we have the Lexus studio. It's a program for processing
• 02:00 - 02:30 point clouds. And on the right side, Liix Cyber Color. It's a software which creates photorealistic 3D models um based on the 3D gion splitting. And therefore I got a video which shows our test field for the indoor. And yeah here we can see we can switch from uh to the point cloud and we can
• 02:30 - 03:00 move freely in this model and yeah I can say that the X grid uh they come super famous for their I think there was one of the first scanner who have a very good caution splitting implementation. And I think the the system runs very uh famous for this. But of course the big one is the real uh measurement system not only an optical or the virtual system. So and then let us start with the yeah measure measurement uh options from the system. Yeah it was my task um
• 03:00 - 03:30 to check the geometry accuracy of the L2 Pro and for that I did some accuracy testing for indoor and outdoor applications. The test field was a hole in the HTV driftton. So the hall that we see uh in the previous video. The first task was to check um the change in room distances after processing with and without grow control points. The second
• 03:30 - 04:00 task was our comparison of the point cloud of the L2 with the point cloud of an terrestrial scanner, the fiber focus. And the last task was um the accuracy verification in auto settings. Um here we can see our test field and its extensions. The hall is also used to calibrate laser scanners. That means it's got many control point markers with highly precise coordinates. And you can see the crowd
• 04:00 - 04:30 control points which were used to reference the scans and the local local coordinate systems. These are here these four green points and maybe maybe it's important to say that this is a very u stable uh facility. So it's a big concrete facility. I think original they developed this to I think they would the basic idea in beginning was to put there some locomotives inside. Oh, and then think then later they don't have these various stuff in the university and then
• 04:30 - 05:00 they make it then for the mechanical engineers actually they're using this and the survey occupied uh the whole structure to put a lot of markers inside and and this building is super stable against vibration and um also I think the climate control is also pretty good inside of this building. Yeah. So let's come to the aim of my study. It was about testing whether the geometry of the point cloud is improved by using ground control points and a comparison
• 05:00 - 05:30 with the terrestrial point cloud of the fiber focus and also the verification of outdoor accuracy. My process was the data acquisition with the L2 Pro, the data processing in Lexus Studio um with and without crowd control points and the data evaluation. Yeah, here we can see um it's about the distance comparison um
• 05:30 - 06:00 between the control points. Here in the upper right corner, we can see where um control points are located. I use the control points from Favo and I calculated the spatial distances between the control points um using the given coordinates. I then carried out measurements in point cap by creating sectional views, measuring the middle of
• 06:00 - 06:30 the control points and calculating the distances and uh a total of 15 differences uh were used between the control points and the distances were between 2.8 m and 40 m. And yeah here on the bottom in the middle column of the chart we can see the value of the deviation from the target distances of the process point cloud without the influence of the ground control points. Um we can see
• 06:30 - 07:00 here 9.2 2 millmters. And on the right side, you can see the difference to the point cloud processed on the basis of grout control points. And there we have 5.3 mm. And the result is that we improved um by using grout control points um about 4 mm. Okay. That's interesting uh that we have an improvement from 4 mm and actual also
• 07:00 - 07:30 the absolute value from 5 mm and 9 mm are super good. Um I haven't questioned if I look on the picture there's also some control points. There's market under the roof. Is was there also sheer bolts there in the middle here from the picture? They are directly under the roof placed in the hole like this one and the other ones. Yeah, we got uh some up here and yeah they are placed everywhere in the hole. Okay. What is the hard work from the university to place them there?
• 07:30 - 08:00 So I think you need some elevator to touch the points. So I guess yeah I think so I think audience can see that it's a very good test field. So not only some points on the sides or everywhere. So so let's come to the software cloud compare. It's a open source software um for processing point clouds and I use this software for comparison of the terrestrial scan data focus and the L2 and for this purpose I put both point clouds on top of each other.
• 08:00 - 08:30 Then I use the function cloud to cloud distance tense which makes differences visually easily recognizable and yeah the reference point cloud was the point cloud of the car focus and the compared of the L2 and the compared one that's um the one which the distances are calculated. So cloud compare calculates the distances of the individual points relative to the reference cloud and the generated scala field is placed in this cloud. And here
• 08:30 - 09:00 on the right side you can see the color scale with the distances. And here's a video which you can see the coloring of the distances to the reference cloud. And here we can see how um the distances goes
• 09:00 - 09:30 up and we can see that most of the differences are in the range of 3 cm. So um nearly the whole hole is uh colored in blue. And we have nearly zero green and orange
• 09:30 - 10:00 areas and here when it comes to the end we can see some red areas. Um, I think that are areas here like we can see that's a door or there's I think um yeah movable objects such as a crane or something like that and the point cloud of the fava pocus was taken I think more
• 10:00 - 10:30 than one year ago. So in the hall there can be changes here. So here we have um some other views from the side, the top and the bottom. Yeah, I think it's the side view. You see there's these red areas of the doors and the crane area and in the bottom we check this areas out. But in the inside there are I think we have shown this on the video. There are some mobile what
• 10:30 - 11:00 you will say them some cells where robots are placed and we imagine that we uh scan there to any fences or like windows. What what was there like I think they have fixed um walls and inside of the walls are some glasses. So I think we can see that the red areas on the bottom line coming from these objects where we scanning through other materials but we can see everything is below this 3 cm would exit offer in their tech sheet. Yeah, that's
• 11:00 - 11:30 right. So I also did a more detailed color scale and yeah here we can see how it's done. So I think yeah there were steps. Yeah I think I can explain it. So I think we then we say okay we are under 3 cm everything after our first check of the data and then say oh let us do it new because uh a few um manufacturers
• 11:30 - 12:00 offering and say we have slam scanner with 5 mm accuracy say let us take this if we can achieve 5 mm and then we take the next step and see then from 5 mm to 1.5 because we in Europe we have often the rule there's a 3D uh model or a 2D uh drawing should be better as 2 cm. So if you can then spend maximum 1.5 cm then we have 5 mm for generalization in drawings and then
• 12:00 - 12:30 finally you can take from these original point cloud in 3D model this was the shift what they have as 5 mm what the best say they have 1.5 this next step and then we go further the next steps so 3 cm is what um offer to the client so yeah let us move forward I'm very interested yeah I also got a where we can see how it's
• 12:30 - 13:00 colored. And we can see most of the areas are colored in blue and in green. And here we got some um yellow. Maybe we can skip a little. Yeah. And we can see um yeah,
• 13:00 - 13:30 mostly of the area are colored in blue and green. And that means that the hall has only deviations. The most of the hole has only deviations up to 1.5 cm. And that's very interesting. I think uh maybe let's move to the next slice. So we can see it also. The blue means uh below 5 mm. There's some many areas there under 5 mm. Then we have all the greenish part what is up to 1.5 and then
• 13:30 - 14:00 we have a little bit interesting here in the side view the small area with the yellow part but yellow can be from 1.6 cm to 3 cm. So that that can be super close to these 1.5 that we actually not can say to people. And then we see the wet areas like we saw before where where we have these movable objects inside or maybe also in some area there's a super shiny uh yeah air channels inside. So where we see more Yeah, that's
• 14:00 - 14:30 right. So let's move on to point cap. Um, point cap is software that is specialized in um, sectional fuse and I overlaid the point clouds from the far focus and the L2 again and I generated cross-sections and with the image resolution of 5 mm and here I did some
• 14:30 - 15:00 measurements of the highest deviation. and did some screenshots which we can see here and most of the hall shows minimal deviations so which wasn't even I I wasn't able to measure them because it looks like one line and here are the highest deviations and you can see here the highest deviation is up to 2.9 cm and so we got the requirement of 3 cm absolute positional accuracy
• 15:00 - 15:30 um Yeah. Okay. Perfect. So, I think we can see in the below picture there's one area. I can maybe task this here with my mouse. I don't know if this this uh captured now. So, there's one where you can see really a double line. I imagine that is maybe an area where we have um maybe shiny material because if you see the rest of the data set, it looks more similar. And I think if you have very shiny objects, I think that's not the best preference for the Sai Sensor
• 15:30 - 16:00 systems. So I think uh maybe the this is the reason why we see a little bit double line on this area but the rest looks super good. Yeah. And I also analyzed the height deviation in point cap and um they were not uh noticeable. So not even in the red mark area. Maybe if you watch the first video that we um recorded um which was about
• 16:00 - 16:30 the K1 the little scanner from X squids there we got some height deviations up to 8 cm in this red marked area and therefore I did um some sectional um fuse in point cap to measure if we got deviations and we can see that um not not um even in the problematic area we got deviations in the height. We can see
• 16:30 - 17:00 it's one line and so I also equipped the hall um with reflective targets here. These are these targets and the points outlined in green were included in the processing while those outlined uh in red were used as checkpoints. And Liix Studio features a function that automatically detects the reflective targets and compares them with the
• 17:00 - 17:30 original coordinates. And the results show a horizontal accuracy of 1.4 4 cm and a vertical accuracy of 1.8 cm what is a really good result. Yeah. So let's move to the auto accuracy analysis and yeah maybe let me tell you something. I was super interested in this. So because uh people are starting now using slam scanner for topographic survey and we want figure out what we
• 17:30 - 18:00 can really achieve by doing this outdoor. So, and that was my idea and so let us start. Yeah. And for that I also used the reflective targets and here we can see two screenshots from the outdoor area of the heart. And on the left side we can see which crowd control points um were used for the red loop here. That's the red loop. And the aim of the study was to
• 18:00 - 18:30 determine how point cloud accuracy behaves under different configurations of GCP distribution. And therefore I also used um here these reflective targets. And we got nine GCPS on our red loop. And the length of the loop is about 600 m. Here we got the test one and therefore the GCP distance was under 100 m. That's also uh the requirement from Xcripts to achieve good
• 18:30 - 19:00 data and yeah the GCPs used in the non-rigid processing um were this one and the checkpoint was the point uh 93 which we can see here and it's located about I would say 20 m far away from the um GCP 164 four and we can see that um the checkpoint and the
• 19:00 - 19:30 RMSSE of the other GCPs are all under the 3 cm um absolute accuracy. Yeah. And and uh for the m metive we take the non-richid. Non-rigid means that we give the software the opportunity to scratch the point clouds on the control points so as that they can they can it's not rigged so it's not fixed so as they can move it on the right position so that that's where they say we can get better results out for the points between the reference
• 19:30 - 20:00 points here the test uh two therefore we got the GCP distance uh above 100 meter and here on the right side we can see um that I only used four GCPs for the processing, but we can see um that they are um they got a good placement. And here are the checkpoints. We got uh five checkpoints in this test.
• 20:00 - 20:30 And we can see that the RMSE of the checkpoints and all the GCPs here uh are just minimal higher than 3 cm. So we also got great results. Yeah. But but I see in the horizontal accuracy we we we going a little bit higher from 2.4 four to three and for the checkpoint I think and but in the high accuracy they will jump from a half millimeter to 1.5 cm so as to 50 mm is three times more
• 20:30 - 21:00 and by the other ones is only 30% more so I think there is something that we can I don't know you can say that but from the points is that the high accuracy definitely increase higher so I think we see it also in the vertical accuracy so but I don't know this is in this is only now our view here it's in general view So uh it's uh pretty interesting. So let's come to the last test. Uh it was about a poor GCP distribution. And
• 21:00 - 21:30 uh so that's how it shouldn't be done. Um we can see I only used the four GCPs here on the top. These are these four and we got all four five um checkpoints. And we can see that the RMS E of the checkpoints in horizontal accuracy is nearly um 6 cm and in height we got um really bad results. So don't place them like that. No. Yeah. But this is a little bit this is like what I see before in the other
• 21:30 - 22:00 picture that's now shows the same. So it it you have a bad distribution of your of your GCPS that the vertical accuracy will be much more affected as the horizontal accuracy. That's not now my my if I see this data but I think we we only check some data so but that would be my my short uh take away from this but I I don't know it's 100% uh science proofed but this is dangerous though if
• 22:00 - 22:30 you're speaking about topographic survey I think definitely if you're doing street uh I think construction guy can live better with horizontal deviations as with vertical so as I don't do that Yeah, but we could um say when we do it as expert says, we get good results also. Yeah, definitely. Everything what we showing here in the end is not what the uh manufacturers provided. That's
• 22:30 - 23:00 our testing. Yeah. More practical testing. So, we know people like cheating. So, doing less than so because time pressure. So, don't cheat. No, don't do this. So, let's come to the conclusion of the Lixel A2 Pro. Um, the manufacturers's accuracy requirement of 3 cm can be met and we got improved accuracy through the use of crowd control points and strength of the
• 23:00 - 23:30 scanner are the great absolute accuracy. We got simple operation, the efficiency and the presentation of the data. Okay. Yeah, that's very good. I I see from my point is what we saw also it's like uh um the need for GCPs to get high accuracy out what but what I really like on on the newest generation of the slam scanner if if I know we starting super early with the first geoslam system so the first development for practical use
• 23:30 - 24:00 of slim technology with the hokuyo scanner I think to this point the the point quality was not um good enough to make any modelings or uh from that and also the camera resolution to this time. Okay, we speaking about six seven years in the past. So was not good enough for do many task of capturing buildings and topographical survey we had other use cases for for this system. So but now with the newest generation and I see
• 24:00 - 24:30 this one uh I think there's complete change how you can use the system. So I think you definitely can capture buildings and you can also capture topographical areas if you use uh then ground control points you can achieve a good accuracy I would say. So as conclusion am I right? Yeah that's right. Yeah I think uh we don't have to be complicated this to the end. So uh maybe to the audience don't forget uh leave a comment if you like to know more. So and
• 24:30 - 25:00 uh subscribe to the channel uh and give a thumb up if you like what Jiannis is doing for you. He spent a lot of time in that. So, and I can tell you uh I got I was on this video IG conference in Prague and there I met some of our YouTube uh uh channel follower and they're speaking about our other um um um accuracy video from the K1. So, and they say they like it and they say that's very professional done. So it's
• 25:00 - 25:30 not uh I don't they see that we really spend a little bit knowledge and try to make it most science scientistic at possible for what we can spend a time for doing that I'm right yeah yeah sure that's I think even you speaking with with your professor about our first results so and make a little cross check with him yeah okay so yeah then to the audience uh very nice and I can tell you once we are not finished yet. So we will bring
• 25:30 - 26:00 more even that you can see on the picture we have this uh instrument here this system have also an integrated GPS what we not was able to using yet so uh I think the actual there is for Europe uh or for Germany their complete transformation system and geo system uh wasn't ready to the point where we check the system so and uh I think that's maybe something for upcoming testing using then the GPS integration and check
• 26:00 - 26:30 out what uh how GPS integration can help us to working better with slam technology. Nian, you are happy to do the to do it. So yeah, I'm happy. We will move on and uh yeah, maybe we'll get some new videos about the RTK. Yeah. Yeah. Okay, cool. So then I say bye-bye. Also, don't forget subscribe to the channel, leave a thumb up. I say bye-bye, Janice, and thanks for your good work. Thank you. Okay. [Music]