Drone Mapping Guide | Part 1
ฝัง
- เผยแพร่เมื่อ 2 ส.ค. 2024
- This is the first part of the tutorial series covering drone mapping in RealityCapture. In this one, we’ll go over importing the images, aligning them, and optimizing the alignment with the ground control points. We’ll take a look at how to deal with the so-called bowl effect, how to deal with inaccurate geolocation of the drone images, and what to do when ground control points and images have different vertical systems.
Timestamps:
0:00 - Intro
0:32 - Adjusting settings
1:45 - Preventing the bowl effect
2:35 - Importing images
3:05 - Aligning images
3: 27 - Importing ground control points
4:52 -Different vertical systems
5:27 - Image suggestions
7:17 - Confirming image suggestions
8:20 - Importing control point measurements
10:29 - Disabling inaccurate camera priors
10:50 - Update vs. Alignment
12:02 - Outro
Part 2: • Drone Mapping Guide | ...
Part 3: • Drone Mapping Guide | ...
Part 4: • Drone Mapping Guide | ...
Download RealityCapture: bit.ly/2sotr7h
Try it yourself with free sample datasets: bit.ly/3cRBfQp
Follow us on:
Instagram / capturingrealitycom
LinkedIn / capturing-reality
Facebook / capturingrealitycom
Twitter / realitycapture_ - วิทยาศาสตร์และเทคโนโลยี
![เกิดใหม่ทั้งทีก็เป็นสไลม์ไปซะแล้ว ซีซั่น 3 - ตอนที่ 65 [ซับไทย]](http://i.ytimg.com/vi/4EsinscQCfs/mqdefault.jpg)
@Jakub - Thank you so very much for this marvelous program Reality Capture and most particularly for this excellent video series. But more than all of that, you and anybody else that's tied to your company, Capturing Reality, are to be highly commended for the business model that you've adopted and how licensing is approached. Finally, a world-class product predicated on the most fair and reasonable fee structuring that I've ever seen. Well done!
Thank you very much, i am Using Reality Caputre for Drone "Mapping" more like Housescanning for like 2-3 Years now and still learned alot.
Ďakujem veľmi pekne Jakub , za každé video. Palec hore
Great Videos! Thanks!
@Jakub - I am having trouble aligning the images - I have done a vertical facade survey of some silos with a 65% photo overlap. When i insert the images and align I seem to get lots of small components that dont make sense. It seems either RC isnt placing the images in the correct GPS locations. Is there a way to fix this? Would uploading the flight logs help at all?
Can you contact our support with this issue?
Great work Jakob, very informative and professional tutorial. Which capture app would you recommend based on your experience, Drone Deploy, Pix 4D Capture, Map Pilot Pro, ESRI Site Scan LE or the paid Drone Link? Is the Terrain awareness important to vary altitude and keep the flight height parallel to ground and keep a consistent GSD? Is a good rule of thumb to keep the GSD typically below 20mm by selecting appropriate flight level? Many thanks.
Dear Lee, we don't recommend concrete application. RealityCapture works with images and it is on your preferences which is best for you. It depends on your work and what you need to obtain. But it is better to follow the terrain because of the same GSD and non changing overlap. Regarding to GSD, it also depends of your needs and what for you are doing that project.
Hello, Thank you for your tutorials. I have a question. In Drone Imagery + Ground control points of free sample datasets, Control points' Altitude is Higher than Images' Alititude. Is these Ground Controls correct?
Hi yes, the control points are correct. Problem is with the image height, where DJI drones use relative height, not the actual one (forum.dji.com/thread-165794-1-1.html). So it is possible in some cases, that there are these differences.
Great set of videos, thank you for going into so much detail. I've a question about the number of GCPs to use. I'm mapping an earthworks site which is around 14acres and am using a total station to set some GCPs throughout the site. On my last model I used 18 GCPs. Is there any disadvantage to using so many, other than taking more time to confirm locations in the images?
No, it is even better. If you have some many of them, you can some use as check points to check the actual quality of the model.
Hi, could you please pass the raid data: altitude, angle and image overlap.
Hello Jakub, Thank you for your tutorials. I would like to ask, that: Is it possible to sort Control points by name? I chceck accuracy my 3D models and I have more than 100 control points and I have problem with orientation in my list. Thank you for your answer Stano
Hi Stano, this is not possible, but we have a feature request about this. I will add your vote there.
@@CapturingReality Hello Jakub, If I can, i have one more question. I´m researching created 3D model without gps, total station and scanner. I would like to define scale of model only with reference distance and find optimal position of scale. Is it possible to define only reference distance and than control accuracy of model with masured points with total station? Because when I don´t choose some ground control point my model gave me to wrong positition and accuracy didn´t wrote true accuracy. When I choosen all my measured points with total station, defined points like ground test it is not moved 3d model to optimal position for minimal errors? Or do you have some idea for check accuracy when I want to define only reference distance? Thank you Stano
@@StanislavStehel Yes, it is possible to scale your model just with the distance. It is better to use more scales (it also depends on the size of the model) and you need to use correct distance (not too short). Are you using drone dataset? When you are using total station, you can also use these points as GCPs (it is better, as they have better precision as RTK GNSS). Ground test is just for comparison, these points don't have the influence on created model. Best way is use some coordinate system or local coordinate system to check the accuracy.
How do I get/create or where do I find the ground control points file?
You can create them manually (Alignment tab/Optional/Add Control Points) and set the points as Ground control or you can import those as TXT file under Workflow tab/Import & Metadata/Ground control).
Do you have this in pdf format too, other than just the videos? I would appreciate that very much.
Hi David, unfortunately we don't have such PDFs. But there are similar tutorials: support.capturingreality.com/hc/en-us/articles/360001577032-How-to-geo-reference-the-scene-in-RealityCapture-using-ground-control-points-faster, support.capturingreality.com/hc/en-us/articles/115002328751-Orthographic-projections-in-RealityCapture
Great Video. I am currently working on making an underwater model of a deep-sea coral garden. I have the lat, long, and altitude of the ROV that took the images. Is it better to import a flight log or ground control points? Or a combination of the two to accurately georeference and provide scale?
It depends on the precision of the data. The best way is to have both (as you can check also the precision of the results much easily).
@@CapturingReality I understand. Thank you.
One last thing. In the context of just ground control points. Is it necessary to add GCPs to all the photos in the model or just specific areas within the model? I am exporting the scaled dense point cloud into cloudcompare and ArcGIS problems as a DEM.
@@ignacio_the_coralbiologist The GCPs should be placed around your captured area evenly. It is not necessary to add them on each image. The number depends on the used dataset, but basically it the point is measured/placed/added on 10 images, it should be enough for most cases.
@@CapturingReality thank you. I appreciate your insightful tips.
I went to your website and because it didn't;t recognize my email as a 'business account" it wouldn't let me request a demo ... so .. I'm moving on to the next option ... shame, it looked viable
Hello, double grid with what degree of oblique angle do you recommend? I typically fly nadir. Is this wrong?
Will it be orthorectified after processing with obliques?
RealityCapture supports true ortho projections of textured models and image mosaicing. For the true ortho, double grid flight with obliques is better. For image mosaicing, nadir shots are better, especially when they are projected on the digital terrain model. Remember that you can use both obliques and nadir shots in a single project and then disable and enable groups of images for specific processing steps.
No, it is definitely not wrong. It depends on the use case. Double grid flight was recommended as one of the options for mitigating the bowl effect. This project used a single grid flight but was corrected with the help of ground control points.
How do you align drone captured images with hand held camera images?
The same way how any other images. If there is a good overlap they will be aligned, if the images are georeferenced, they will be aligned. When not, you can use control points to merge two separate components. The principles could be take from this tutorial: th-cam.com/video/kKA-KxT25KA/w-d-xo.html
what case the "A depth-map is corrupted or misssing" error when reconstruction?
Hi, please check this post: forums.unrealengine.com/t/a-depth-map-for-filtering-is-corrupted/1698231
@capturing reality How does RC handle Check Points? In accurate mapping, processing without CPs is just pretty pictures. Maybe it's in one of the next vids? Thanks!
Hi Dave, you can choose some of the Control points as Check points and then find out the accuracy of created model. In RC it is called Ground test. In this video the control points are in the Importing control points measurements part.
@Jakub - Many thanks for this tutorial! I'm totaly new to this RC and have some questions about differences between the files you use in your tutorial and the online datas provided on RC website:
1 - Coordinates system:
You use epsg: 5514, and on the website description about coordinates are separated in horizontal epsg: 4258 and heights in Baltic vertical epsg: 8357.
How should i deal with those informations in the RC settings? Am i supposed to use the epsg from the country i work from or the epsg the pictures were taken from?
2 - Group calibration Exif:
All images from downlaoded file are the same focal (e.g 24mm). Is it necessary to change GCExif to "Yes"?
3 - Distortion model:
How do you know what DModel to choose? Is it relative to the drone/optics constructor?
4 - GCPs:
The downloaded file is already provided with Ground Control Points going from 101 up to 203, but there is no negative values like in your tutorial. If i import GC with File Format X, Y, Z i get an error from RC, so i assumed they were more lat, long and alt coordinates and it worked. But i had to manually adapt the images suggestions because they were all above the picture in the 3d viewport. Everything worked fine until the .csv step at the end of wich i got an error msg from RC. How should i use those informations? Also the .csv infos are provided with the downloaded files, i don't know where to go with those, like i'm confused!
5 - Readme file:
It says we should change the path to the source folders in the SetVariables.bat as follow:
* RealityCaptureExe - path to installation folder RC.exe is stored.
*RootFolder - path to all source folder.
How do i do that? (cmd prompt, but don't know how to do it properly)
That's a lot of questions, so in advance thank you for your answers Jakub!
Hi Vincent, I will try to answer your questions:
1. The dataset which you can download from our site is another than used in this workflow. Regarding to coordinate system, you can set Project and Output coordinate systems in Application settings )(they could be different). Also, if you are importing each files (like GCPs you can set their own coordinate system). You can set these systems as you need or to in which system are the data. In your case you can set 5514 for project and output coordinate system, but 4258 when you will import the GCPs.
2. It is not necessary but you can do this. It is better do do when you have different focal lengths.
3. For most cases is OK to use pre defined distortion model. When you have fish eye optic, then you can use Division.
4. Yes, GCPs should be in Lat/Long (4258). It is quite usual thing to manually place the GCPs over images, as drone data are not always in correct heights. Usually they are measured from the take off height. Did you realign the component after GCP's placing? Also, you should set No in Use camera priors for georeferencing under Alignment settings in this case.
5. You can open that BAT file in Notepad++ for example and change these paths there.
@@CapturingReality - Thank you Jakub, i'll try this later this afternoon and let you know how it goes!
Hello Jakub, i did it with the script, i used Visual Studio Code to open files. Everything went well, while during the process RC crashed right after the texturing operation. Don't know why, might be because i'm only 16Gb ram?
I will try to do it step by step today, like you do in the tutorial instead of running the script, and let you know how it goes. Thank you so much again, have a nice day Jakub!
@@vincent4468 Hi, did you get some error during/before crash? Were you using the setting from script or did you change some of them?
No Jakub, no error before but a crash during the execution of the script right after Texturing step and i did not change anything in the script. I will try again and send a crash report if it helps and if possible.
Welcome to the learning curve
At the 9th minute and 19th second of the video, you import the CSV file containing the image coordinates of the images associated with the points. I also followed the steps you showed. At the end of the steps I took:
Operation warning
The file contains 42 images that are not in the current scene. Please check the console for the full list of images. [error: 18002]
I get the warning. I don't see the additional points in the rest of the video. what should I do?
Hi, this is just a warning message that there are some images in CSV file which are not included in your project. You can continue ordinary in the process.
@@CapturingReality Thank you for taking into consideration the question I asked, being interested in it, and answering the question I asked.
Hello. Could you make a tutorial using a combination af rtk or ppk georeferenced images combined with GCPs and explain how to set the weights?
Hi Nikou, thank you for your ideas. The principle is the same. You can set the higher weight for data, which are measured more precise. I added this idea to our future tutorial topics list.
@@CapturingReality Hello, I wanted to follow up on this. Have you updated a video on RTK georeferenced images. I am unsure how to process after aligning the photos without ground control points.
@@stephanieramos8464 Hi, the video wasn't updated. You just need to set the precisions for the images in the Alignment settings (Camera prior settings).
Are the images you used in the demo available to work with.
Not these specific images, but there is another drone dataset with ground control points available in our sample datasets on this link: www.capturingreality.com/SampleDatasets
@@CapturingReality I've downloaded the data set you are referring to and tried using it to measure area and volume. The logs just aren't as good an example as the sand would be. Thanks anyway.
During alignment my CPU is at 100% and the GPU is at 0%.
Should RC use the GPU for alignment?
It's an Nvidia Geforce GTX 1660 TI
Thanks
GPU is not used during alignment process.
Will the dataset for this video ever be published, please?
Not this one, but you can find other datasets here: www.capturingreality.com/sample-datasets
@@CapturingReality Thank you.
The 'Suggest Measurements' feature isn't the smartest compared to Metashape, but the workflow for adjusting gcp's with images is still rather efficient.
10:29 *Disabling inaccurate camera priors*
With Ctrl+D and an immediate Ctrl+A what is selected is actually just part of all images; I think it's either all the just prior manually georeferenced images or all images that still aren't. Is this intended or should all images actually be selected?
Hi, using CTRL+A all images in that component should be selected.
@@CapturingReality Hi, thanks for replying this fast. Just noticed that, in my case, it was just because I had _multiple components_ and _RealityCapture_ just functions that way that via pressing Ctrl+A it only selects *all the images for the current active component* which is all that is needed of course :D Nice!
@@mpiauro3958 You're welcome. Also, if you will have open images in 1Ds view and will use CTRL+A, it will select all images in the project.
This tutorial feels lacking of an introduction. Can I use this technique with any drone (such as Mavic 3) or does it need some special GPS or "ground control"?
Hi, there is also the introduction video: th-cam.com/video/j62Xgb2EYNg/w-d-xo.html. The technique can be used by any drone. If you don't have GPCs, you can skip that step.
Can yu make an RTK processing option?
It is possible using flightlog or if the position information is baked to the used images.
@@CapturingReality can you make a tutorial?
@@mariohorvat9330 th-cam.com/video/oMnZFOOK7LQ/w-d-xo.html - RTK/PPK data support, it is an older tutorial, but principles are the same
@@CapturingReality ty
Is there a way to export the camera to 3ds max?
Cameras can be exported as a part of the model.
@CapturingReality thanks. I'm using 3dsmax. I found I can get a individual camera per photo with fbx. Is there a way or a tool to have that be an animated camera instead for tracking?
@@mikegentile13 What do you mean with animated camera?
@CapturingReality i am trying to use reality capture to do a tracking shot in 3dsmax. The footage I'm tracking is 200 frames. RC generates a nice model and 200 individual cameras. I would like to know if it can generate an animated camera
@@mikegentile13 In RC you can create an animation (th-cam.com/video/RLgFpbmJ4WA/w-d-xo.html). But that will result as a video, the cameras won't be exported from that.
Wish you would create a set of mapping videos without ground control points. Honestly most people don't use them
Hi Darrin, thank you for your idea. In that case it is basically the same, but except GCPs you will use flight log. Do you have RTK drone?
@@CapturingReality oh no I don't have gcps or rtk. I have always used drone deploy in the past but I have just started using reality capture and I love all the customization and In my opinion the maps and models created with reality capture are so much better than drone deploy. Most of the mapping I've done commercially are only concerned with relative accuracy and not so much absolute. I love the tutorials but they were so indepth with incorporating gcps it was hard to follow. That is why I was suggesting possibly another set but with out the gcps. Thank you so much for the quick response. Looking forward to learning more about the software.
then it is not maping but creating pretty pictures
@@CapturingReality If your using an rtk drone and GCPs would you still set camera priors to unknown and realign?
@@CAPTURECI You can do that. Then GCPs are used for georeferencing.
kenya and beyond
I’m on a watch list now game on feds
32k views and only 80 comments? why are you deleting comments mate?
Not everyone is continuing watching,, nor replying
Why you spread misinformation?
I love the federal government and I am so thankful to the heroes that browse the web protecting me from stuff ❤❤
hey glowies, come and investigate me
lol beat me to it