Most of photo mapping online services and several softwares as Photoscan use Self-Calibration method for modelling internal geometry of cameras. While this approach is simple and practical, don´t provides the best results because the low convergence and redundancy of a tipical aerial survey.

Atachhed to this blog there is an example of a full field calibration of a Sony Nex 7 with a Sigma 30mm F2.8 (tipical setup in Bramor UAV). Thanks to the large calibration field, sharp pictures were taken with the same focus setup that used when flying (The focussing ring is blocked in Bramor).  If you are lucky enough of own this camera/lens maybe you can try this calibration certificate. All right, for sure, your camera will be slightly different to this, but I dare say that this modelling will be more accurate than that obtained by self-calibration.


E-mail me when people leave their comments –

You need to be a member of diydrones to add comments!

Join diydrones


  • I agree with Enoc, proper pre-calibration is still the key for a precise work. It might not be important if you are using poor zoom cameras but for instance with modern mirror-less cameras with fixed lenses it definitely makes sense. With the recent progress in the field of inertial navigation and precise positioning, the trend of UAVs is going to follow classical airborne photogrammetry with the direct-georeferencicg approach. The precise knowledge of internal parameters is thus crucial.     

  • Without convergency is not possible a realistic modelization of radial distortion in all area of sensor. I think best methodology is start the automatic tie points referencing from an static calibration, and then to do a refinement throught a selfcalibration for account slight changes in camera, temperature, etc.
  • T3

    I don't believe in static pre-calibration simply because each landing  of the UAV is potentially changing lens geometry, the same applies to air temperature.

    I think online calibration is superior exactly fo rthe results mentioned, whil all this calibration certificates are merely artifact and lobbying of old style aerial surveyors, when cameras were unique, very heavy and operating in relatively trouble-free conditions (manned airplane)..

  • Martin, targets are identified automatically. Most of precessing is done with Photomodeler. I stay on the ground during all imaging capture. I agree totally with you that a 3D calibration field ( non planar) and taking images flying would be better. However actual setup is convenient and enought for get very good internal parameters with deviations 2 digits magnitude lower.
  • Ho do you identify the targets? Manually or automatically?  Could you show us the positions of the cameras- did you stay only on the ground during the image capturing or did you also use something like a telescopic platform for taking pictures from above? To get better results, the calibration field should not be flat and the images should be taken under conditions that are similar to the field samples- to fly above the calibration field with the drone

This reply was deleted.