RealityCapture tutorial: Merging exteriors and interiors of buildings
ฝัง
- เผยแพร่เมื่อ 11 ต.ค. 2020
- Learn how to merge the interior and exterior of a 3D scanned building.
The data capture in this project was acquired with a drone and DSLR camera.
00:32 How to capture images
03:28 Automatic alignment of the images
06:19 Inspection of the alignment
08:06 Preparing a dataset without overlapping images
10:26 Placement of the control points
16:15 Alignment of the interior and exterior using control points
Download RealityCapture (PPI) for FREE here: bit.ly/2sotr7h
Download the dataset from the video to follow along www.capturingreality.com/down...
Follow us on:
Instagram / capturingrealitycom
LinkedIn / capturing-reality
Facebook / capturingrealitycom
Twitter / realitycapture_ - วิทยาศาสตร์และเทคโนโลยี
Very informative, thank you!! most of the photogrammetry tutorials only show you the RC settings you should use.. for me the info how to capture the photos is essential !!!
This is an awesome video! Thank you!
So I used this to fix a plant of malbec where I was getting 2 components of the plant. It was hard because I didn't use markers on the scene when I took the photos and I was unable to go back to the vineyard to make a new set. Thanks a lot.
I live near the church too :D
Pozdravujem zo Slovenska!
Výborné tutoriály
Sooooo helpful, THX~
really great tutorial, tomorrow I will try to add to my 3D model of the historic church also its interior
Helpful video 💗 i hope add videos for underwater photogrammetry also
Second Nikou's suggestion, have been attempting to create high resolution textures and reprojecting them onto simplified models but can never reproduce the quality of the original texture on the simplified model.
Thank you for sharing such a detailed tutorial. It helps me a lot for understanding the workflow. I have a question regard camera on a pole. Were you referring as in camera on a tripod or is it something else like a pole with wheel underneath. Because you mentioned that you would set it to interval timer shooting. Then it can be challenging to avoid blurriness. I also tried to shoot around the small village in my city. I shoot with cellphone without the tripod so it was difficult to maintain the height and avoid blurriness. if you can please provide some tips to capture image without blurriness. Thanks!
Basically, you need to have set a quicker shutter and try to be calm during capturing without movement. I think it is mainly on practice to find the best workflow for you. Also this could be helpful: th-cam.com/video/9e_NLp_FaUk/w-d-xo.html
Amazing quality of work and thanks for the detailed explanation. I have a begginer question for you: as i understand pictures that are taken by the dslr camera are not georeferencied. How is it possible to use them in photogrammetry without geo information. By drone I understand that a gps is integrated in the hardware , but for indoor gps is not detected.
Thanks again!
RC doesn't need and geoinformation. It uses photos to find features. It then matches them between images, and based on this, it can calculate the position of the features and cameras in 3D space.
question about workflow.....
After the section on viewing the unregistered pics, you state you realigned the images.... Simply bringing up the unregistered pics like at 5:40, improved the alignment when it was done again?
And it looks at this point, a model was built but it was not mentioned.... is that the case, a model was built?
Thanks!
RealityCapture used the 1st alignment as a base and during 2nd alignment it was able to add a couple of unaligned images. The alignment improved on its own during the second run. No user input. The model was built after the second alignment. Each alignment creates a new component without any models.
Hello from Russia! Thank you so much for your videos!
I have a question: in my model there are almost 3,000 photographs and they are correctly executed, but the whole model turned out to be mirrored; can you please tell me how to fix it? Thank you!
Can you please contact our support with this problem?
Hi, So if i understand properly you need to take lots of pictures around the door and eventually add manual tie points.
How do you reach cm level accuracy inside the building ? Do you need a total station to "project" GCPs or markers inside the building ?
Thanks !
Again great work !
Thank you for your kind words. Like I showed In the first case you don't need manual tie points if you have enough photos with high overlap. Reality Capture will align everything in one go. You have to use manual tie points only if the interior and exterior are not connected with photos like I showed in the second scenario. A total station is not necessary to achieve high accuracy. If you want to work just in a local coordinate system you can define distance constraints. Just take measurements with a tape or a handheld laser meter. Measure some distances on site and define the the distance constraints in RealityCapture. I did not show this in this tutorial because distance constraints were covered in another beginner tutorial and I didn't want this tutorial to bee too long and complex. Of course if you want your project to be georeferenced than use a GNSS receiver to measure some GCPs in the exterior and use a total station to measure some GCPs in the interior. It really depends on what you need and what you are going for. Or you can even combine photogrammetry with terrestrial laser scanning in RealityCapture. We have a tutorial series dedicated to this workflow from CyArk on our TH-cam channel.
@@jvanko89 Thank you very much for these detailed explainations. It helps a lot. Looking forward to work with Reality Capture
Nice! Any tips on how to deal with windows, glass, mirrors and other transparent or reflective objects?
These are still a problem for photogrammetry. You can cover them or create some kind of texture on them (by dust for example).
Make a video for achieve best texturing results
Hi Nikou, thank you for the tip, we will add this to the list of topics for the future.
@@CapturingReality Would be indeed really really really appreciated ! :)
Is there a way to have an error report? Like an accuracy report?
Yes, there is option to create reports.
You can find it under Workflow tab/3. Outputs/Reports.
Those reports can be also modified by you: rchelp.capturingreality.com/en-US/appbasics/reports_functions_and_variables.htm
Hello, I will need help... I am creating pezinská baba hillclimb track and i cant merge all the components together. I have aligned all images but it split into 4 components. even i have correct control points, nothing isnt working for me. any solutions?
Hi, are your images georeferenced? If so, you can use Merge georeferenced components option in the advanced alignment settings.
Also, how are your control points placed? Are you using minimally three of them placed over all components on minimally three images for each component?
How would you add new images to exiting component?
You need to import them and then align again, which will create a new component.
How capture inside .take all the posible angle ?
You can watch our another tutorial: th-cam.com/video/9e_NLp_FaUk/w-d-xo.html
Hi! Could RC import 360 pano photos? Thank you. 😊
Hi Lucio, unfortunately, it is not possible. But it is one of the feature requests.
Thank you so much! Have a nice winter holyday!
What's the group image for?
Hi, check this Help article:brchelp.capturingreality.com/en-US/appbasics/camerasettings.htm
can i delete an imported images?, how?, i can't find out how to delete the useless images
Hi, you can't delete unwanted images but you can disable them for other computation. In 1Ds view open Images and if you go over unwanted image, there is X in red square, which will disable the image. You can also select more images and disable them in Selected input window
@@CapturingReality umm okay i knew it, thankyou
Setting directions (ground plane) makes a lot of trouble. The axes are unnamed. It is too easy to make a model upside down, that fails to merge. Could you add simple icons to the axes, at least where is north and top. Yes, I recommend geographical directions instead of left, right, front, back. They are misleading. When we are inside of an object, looking left we see the right side. Using West, we always look to the west, no matter if we are inside or outside.
Hi Abrimaal, thank you for your ideas, I will add them to our feature request database. The axes in RealityaCapture: green - Y, pointing to north in georeferenced datasets, blue - X, pointing to west in georeferenced datasets, red - Z, pointing up. You shouldn't use geographical directions in non georeferenced models, because there is not North, or other sides.
@@CapturingReality Yes, there is no use of geographical directions when we work with objects that can be moved anywhere. This is still better to use RGB colors of axes than XYZ - different in every editor. I didn't know about the colors, thanks