Scan inside and outside a house?

[chris741]

If I were to scan an entire house inside and out. What is the correct workflow?
I tried doing it all in one but the inside image quality is not that good
Is the correct way to scan each room individually and combine it later in blender?

[Bzuco]

For interiors it is good to have good camera, where you can safely increase ISO to 800-1600 if you do not want to use tripod...so full frame DSLR ~18-20Mpix... more Mpix will cause more noise.
You can put some markers on walls which are painted with just color and have no other features.
Cover glass panels/mirrors.
If you can keep the doors open, it helps you grab more feature points on photos from rooms you see through the open doors.
Interiors are overall harder to shoot because not so much space, plenty of furniture, not good lighting etc.

[Alexey Pasumansky]

Hello chris741,

Some recommendations are given in the following article:
https://agisoft.freshdesk.com/support/solutions/articles/31000159101-terrestrial-laser-scanning-data-processing

For this project, we have used aerial data for exterior parts and laser scans mostly for the interior parts of the building. Additionally images from the handheld camera have been added for the interior part, as obviously, it is not enough to calculate the corresponding system between interior and exterior subsets because they do not overlap.

We used terrestrial images (from a handheld digital camera) of the parts that can be seen both from outside and from the inside. For example, photos of the doorways and window holes have been made, while moving from the outside to the building's interior part, thus providing overlap between the external and internal part of the building.

If you get better results scanning the rooms individually, you can try to use markers for the relative referencing of separate rooms and exterior in the same coordinate system. For that you need to measure 3-5 points for each subset, so that all the coordinates are in the same local system. Also using markers in doorways, as Bzuco mentioned above, may be helpful to align the rooms together (when processed as individual chunks) using marker-based chunk alignment approach, providing that markers can be marked in both (or multiple) chunks being aligned.

[chris741]

I only have the Standard Version which is a little bit limited with markers and laser scans
The professional version is too costly with way too many features I will never use

To clarify I dont have an issue with taking pictures and processing them all to get a single object. I have been able to scan the entire outside and a single path with rooms through the inside.

What Im after is the is the workflow of how it should be done in the application
What Im asking is do I just use one chunk and put thousands and thousands of pictures in it? and just process everything all at once? The problem I seem to have is the details of the textures seems a bit lacking inside because it seem to be trying to consistent scale of details making the same detail level on the inside as the outside
Then if I have an issue with one room I have to waste hours for a rebuild of everything.... Not what Id call ideal

or

Can I split it up the rooms and process each individually so they have their own texture. Then combine them all into a single object with multiple texture iages.
Then when I have an issue with on room, I correct it, process the one room and combine everything again
Can something like that be done?
If so how?

Can I use a 360 camera and import multiple spherical images to recreate each room?

[Alexey Pasumansky]

Hello Chris,

I was referring to the article mainly for the image acquisition hints that were used to get overlap between exterior and interior subsets.

If you do not have issues aligning all images from different rooms in the single chunk, then it is better to perform the camera alignment in the single chunk, thus you would have everything in the same scale and wouldn't have a need to align the results of individual room reconstruction later.

Basing on the same alignment you can generate textured mesh for each room by controlling the bounding box size in the Model view before building the mesh and texture. Following this approach you can get several mesh models in the same chunk that are related to different parts of the complete project. Those separate meshes can be exported individually or merged into a single model after texturing.

Spherical panoramas can be processed in Metashape to get textured mesh model, see the related instruction:
https://agisoft.freshdesk.com/support/solutions/articles/31000161406-processing-spherical-panoramas-in-agisoft-metashape
Spherical and cylindrical sensor types for equirectangular panoramas are supported in Standard edition, so the workflow from the article can be followed in Standard edition as well.

[chris741]

What about the mesh detail? I dont need the exterior mesh to be as high as the interior
How do I do the exterior without the interior?

[James]

one possibility would be to disable all interior images while making the mesh of the exterior