Using Total Station coordinates to scale project

[Geooooo]

Hello,
I am pretty much a beginner to photogrammetry and 3DF Zephyr, however I have been making progress by reading documentation online and watching videos.
What I am trying to do now is scale the project using coordinates taken from a total station.
I took pictures around a building to produce the mesh and took coordinates at 4 points using the TS.
I have been able to rotate the mesh based on the coords (I think), but I am not getting accurate measurements.
In this case I made the mesh first and then added control points by the images and then imported a text file with the 3D coords.
I then tried to add the constraints in the SfM phase but it said that there weren`t enough points?
I have read the manual and watched the tutorial video regarding the control points but I must be missing something or maybe I can`t get accurate measurements because of the pictures I took.

I would try to figure it out more myself but as I already activated the trial licence last week I am against the clock to produce some good results, and then hopefully be able to get a perpetual licence!

I would appreciate any help in this matter.

[Andrea Alessi]

Welcome Geooooo,

you generally want to add control points immediatley after doing the sparse point cloud: this way you can use the control point as constraints, so you can run a bundle adjustment. You can still do that afterwards, but you will lose everything that is not a sparse point cloud.

If you just use the control points as constraints to rescale the world, it will do just a rigid transformation.

I'm unsure of what you may be missing exactly - what is exactly the message you get?

A good idea is to check the orientation by placing control points and verifying the reprojection on the epipolar line. This tells you quickly if your project is correctly oriented or if there are bigger issues that you need to tacklet. Regardless, you should certainly be able to import your control points and use them.

If you can share a little bit more details of your workflow I'll be happy to help - or feel free to send everything (zep file and related subdirectory, photos, .txt with the coordinates) to support@3dflow.net and we'll be happy to take a look!

[Geooooo]

Andrea,

Thank you for your quick and helpful reply.
I did not realise that it was best to import the control points directly after creating the sparse cloud.
I redid the project and the measurements are much more accurate this time. Thank you for your help.
I have also emailed you the project, as I would be grateful if you could have a look at it, and maybe suggest any improvements that I could make.
As I said I am pretty much a total beginner so there is still a lot to learn.
Could you also explain to me the meaning of bundle adjustment and "reprojection on the epipolar line"? Pretend that I am really, really stupid (I am).

Thank you again for your help and I will continue to study the available materials.

[Andrea Alessi]

Hi Geooo,

no worries, we're always happy to help!

importing control points right after the sparse cloud is a good practice because then you can do the bundle adjustment. Without doing the BA, you can import them also afterwards. The BA is a minimization procedures that drives the solution using your constraints.

however, doing the BA means that the camera positions will be optimized using your constraints (2D and 3D). If you execute the BA, you change the cameras, hence you get a different result than doing just a rigid transformation, where the constraint isn't used.

I suggest if possible to use photos rather than a video stream for a capture like yours, and if possible, to add different heights when looping on your subject. Narrow, continous paths like yours, may suffer from error accumulation and you want your images to be as clear and crisps as possible. Video streams are great when you ahve to take a dataset very quickly or when you can't take photos for other reasons, but in an urban scenario I strongly suggest taking pictures rather than a video stream.

you can verify the reprojection on the epipolar line when you place control points. If you see the local and global reprojection with values that are too high, there is likely some orientation issue you need to fix. Ideally you want the reprojection value to be below 1 pixel, especially when you're dealing with very low resolution images.

Additionally, I suggest using more photos to place your control point.

4 control points are not a lot in general, but for such a small test dataset can be enough. However, I suggest first improving a bit your capture technique as the photos are very blurred, given they are coming from a video stream. Taking the proper time to acquire the dataset (maybe with the help of a tripod and a better camera) will yield significantly better results.

Here are some videotutorials that I hope will help

https://www.youtube.com/watch?v=E06kgYBftak

https://www.youtube.com/watch?v=nQozgfnKBM8

https://www.youtube.com/watch?v=HZtcJOZOstU

[Geooooo]

Andrea,

thank you again for your helpful reply.
Actually, I thought that I was using pictures in the project. But you are correct, they are stills I took from a video.
I also took pictures at the time, so I will have another go at the project using those.
Thank you for your tips on capture technique, I will keep practising.