Agisoft Metashape

Agisoft Metashape => Camera Calibration => Topic started by: sven09 on March 05, 2017, 03:29:12 PM

Title: Negativ Focal Lengt - Fisheye
Post by: sven09 on March 05, 2017, 03:29:12 PM
Hello,
i am playing with Camera Calibration and iam not shure if the calibration is correct.
I tried a gopro hero 4 silver and a xiaomi yi 2k with different photo sets (5-60) and i always get an negativ focal lenght.
Is this normal for an fisheye camera?

And my second question xiaomi yi offers a "Lens Rectification" mode, which turns off the fisheye effect.
If i use this mode should i use frame or fisheye mode? I am not sure because there is also a light fisheye effect left.
I attached one calibration photo with fisheye off.

Thanks for support
Sven
Title: Re: Negativ Focal Lengt - Fisheye
Post by: Alexey Pasumansky on March 05, 2017, 03:34:38 PM
Hello Sven,

Usually we do not recommend to apply geometrical transformations to the images before processing.

This particular image looks to be still fisheye, but I suggest to process the original images anyway.  If you are still getting negative focal length, please send the .lnz project file and source photos to support@agisoft.com.
Title: Re: Negativ Focal Lengt - Fisheye
Post by: sven09 on March 05, 2017, 06:11:37 PM
Thanks Alexey,
I have tried with standart fisheye images and sometimes its work and sometimes i get negative focal lenght.
I have a photo set with 19 images but i dont know if the result is accurate.
Send you a mail, hope you can help me, maybe my photos are just bad.
Title: Re: Negativ Focal Lengt - Fisheye
Post by: Alexey Pasumansky on March 05, 2017, 06:43:35 PM
Hello Sven,

In your project the focal length seems to have adequate value.

But are these images already modified or these are originals?
Title: Re: Negativ Focal Lengt - Fisheye
Post by: sven09 on March 05, 2017, 06:50:51 PM
The photos in my mail are without modifications.
Should i make more photos to verife my calibration file?
I am not shure how accurate the calibration should be, should i focus on an accurate calibration or arent there big differences?
Title: Re: Negativ Focal Lengt - Fisheye
Post by: Alexey Pasumansky on March 05, 2017, 06:54:53 PM
Hello Sven,

The number of photos is already sufficient. More sharp images will allow to detect the corners of the cells more accurate and should improve the results.

You can also try to take a photos by the same camera of some well textured object or scene and process them in PhotoScan and optimize the results. Then you can compare the results. However, for such test it's better to use the image frame space as effective as possible.
Title: Re: Negativ Focal Lengt - Fisheye
Post by: sven09 on March 05, 2017, 07:16:39 PM
Thank you, quick support on sunday afternoon!
I will try to make more sharp images and compare them with the opimized results photoscan showes me.
If me calibration file is good should i fix the camera calibration or should i still optimize?
Title: Re: Negativ Focal Lengt - Fisheye
Post by: Alexey Pasumansky on March 05, 2017, 08:01:20 PM
Hello Sven,

Usually, when the calibration is reliable, it should be fixed, otherwise, there's no need to use pre-calibrated information, if it needs to be adjusted.

But if the calibration results should be used in real projects, it's necessary to maintain similar conditions, so calibrating with the Lens application in laboratory conditions the camera that should be used for aerial surveys, for example, might be not really good idea.