Meshroom is effective free software for 3D scan

[Jurassi]

Last days I test free open-source Meshroom software and I found it simple (select pictures, then push start) and effective. Detailled modelling require to take pictures near the building (5 to 10m far) which is difficult with trees around... Flightplan's tracks are a nice way to scan huge areas with pictures every 1 second or every 5m (GPS). The photogrammetry process requires to precisely define optical path (real focal lenth, real sensor width) that is not very clear with Anafi : Anafi resolution is 4608×3456 pixel with HDR, maybe 5444×4016 without HDR ? Sensor width is 6.10mm but for JPEG RECT or JPEG WIDE ?

Meshroom creates OBJ files that can be import in Blender software to create 3D renders.

Meshlab software should also be used to optimize meshing.

 

[DutchMountains]

Thanks for sharing - looks interesting

 

[S.C.C.]

Last days I test free open-source Meshroom software and I found it simple (select pictures, then push start) and effective. Detailled modelling require to take pictures near the building (5 to 10m far) which is difficult with trees around... Flightplan's tracks are a nice way to scan huge areas with pictures every 1 second or every 5m (GPS). The photogrammetry process requires to precisely define optical path (real focal lenth, real sensor width) that is not very clear with Anafi : Anafi resolution is 4608×3456 pixel with HDR, maybe 5444×4016 without HDR ? Sensor width is 6.10mm but for JPEG RECT or JPEG WIDE ?

Meshroom creates OBJ files that can be import in Blender software to create 3D renders.

Meshlab software should also be used to optimize meshing.

I'm interested in 2D only aerial photography, but natural landscape is often not geometric enough to produce precise stitching, could this software be of any help ? I don't need a $150/month software...

About optical data, from Anafi's User Manual :

23 mm f/2.4 lens, sensor is 1/2.4', ratio 4:3

Large JPEG & DNG 21MP 5344 x 4016 px (84° x ~63°)

Rectilinear JPEG 16MP 4608 x 3456 px (75.5° x ~57°)

Focal length and aperture do not exactly fit with some basic photographic rules, though (f/16 exposition, etc), maybe it's because the sensor is so small.

 

[robinh]

I hadn't put too much thought into this and assumed that rectilinear and wide angle were different picture shapes. I thought I would just share this in case anyone else doesn't know what I just found out.

• Wide: 21MP (5344x4016) / 4:3 / 84° HFOV
• Rectilinear: 16MP (4608x3456) / 4:3 / 75.5° HFOV

The wide angle uses more of the lens and hence there is distortion around the outside of the picture.
Rectilinear means (of a wide-angle lens) corrected as much as possible, so that straight lines in the subject appear straight in the image. So it uses less of the lens and sensor but is the same 4:3 shape.

So when you're doing scans for photogrammetry you should use the rectilinear setting.

 

[robinh]

I don't think your angle calculations are correct. It looks like you've divided the horizontal angle by the image aspect ratio. Angles aren't proportional to projected sizes. I calculated the angles are:
84 x 68.1
75.5 x 60.3
beta = 2 x atan (3/4 tan alpha/2)

 

[robinh]

I found that Meshroom recognised the ANAFI camera and seems to have the calibration file in its database. I made this spreadsheet which helped me plan my flightpaths. Hopefully people will find it useful.

 

[epysod12]

Hi ,

Thank you for this calculator, it helps me a lot to better manage my flightplans, currently made with Pix4Dcapture. I'm also preparing a script to parse informations contained in the Pix4D JSON file (with creation of a mini-map), but I'm a little confused with the info bellow:

As you can see, image dimensions are 4608*3456, it's ~"OK"~ (we'll see later that this is the tricky point I'm working on), but I don't understand the sensor width!
Why 6.194? The CMOS Image Sensor is the IMX230 from Sony, with the dimensions: 5.99 mm * 4.5 mm.
The only source I find with this resolution is here, on this PDF: Presentation-1.pdf (page 19)

You seem to be comfortable with this kind of calculations, you will surely have some interesting ideas to suggest to me!

If we calculate sensor dimensions, with Pixel Size = 1.12 μm (H) * 1.12 μm (V), we obtain:

JPEG Wide ->	X Resolution: 5344	Y Resolution: 4016	Sensor Width = 5.99 mm	Sensor Height = 4.5 mm
JPEG Rect ->	X Resolution: 4608	Y Resolution: 3456	Sensor Width = ?!?	Sensor Height = ?!?

The Meshroom cameraSensors.db seems to agree with this: [...]cameraSensors.db#L4991

6.194, where do you come from?

To check these measures, we can go to this sites for instance:
Maps Made Easy - Point Estimator (at the end of the page)
Sensor diagonal & sensor ratio

Now I ask my questions explicitly:
_ I attempt to calculate the GSD from the JSON file, with Resolution Image / Focal Length / (optional Sensor Width), is this possible or not? I already search formula on internet, without results... Have you an idea?
_ In this thread: [...].3126/page-3#post-34189, @MustangPhil seems to say that the focal length is not the same for JPEG Rect, that would explain the inconsistencies between the values... Can you confirm with your point of view?

For my part, I'm refining my script in order to distribute it here as soon as possible!

Thank you for your attention,

epysod12

 

[epysod12]

Hello ,

Finally, I keep the formula that I had previously found:
GSD = (( Working Distance * Sensor Width ) / ( Focal Length * Image Width ))

The (Bash) script I'm finishing right now produces an image like this:

It's exciting to see what we can extract from the JSON Pix4D file! To be continued... in another forum thread!

epysod12