Hi,
I'm currently researching on the accuracy of DEM produced from Phantom 4 Pro from different configuration (GCP, altitude, site, and different days of observation).
My question is, do you guys use pre-calibration - as in you have calibrated the camera beforehand by using pinboards, TV, desktop etc - and obtained the camera calibration parameter values?
Or do you guys use the auto-calibration run by Agisoft during the Photo Alignment process?
Appreciate you answer, thank you.
Dear Faiz Arif,
The low-cost UAV Phantom 4 is equipped with a non-metric digital camera (if the FC630 digital camera is applied), and you should expect variations on the focal length and also a large radial distortion (mainly k1).
There are different methods to calibrate your camera trough computer vision algorithms by using a chessboard with known dimensions using OpenCV (https://docs.opencv.org/2.4/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.html), Matlab (http://www.vision.caltech.edu/bouguetj/calib_doc/) or using Agisoft Lens – these approaches are faster and yields reliable IOP, and also are highly automated (decreasing the systematic errors due to the task of measuring GCP on images). Indeed, inserting pre-calibrated Interior Orientation Parameters (IOP) yields better/equal results at a lower cost – in regards to the necessity of surveying GCP to indirect imagery orientation and also to model the lens distortion.
If self-calibration (trough "Optimize camera alignment") is applied, I recommend you to:
- Increase the number of GCP;
- Increase the overlap rate (forward and sidelap) – to generate a large number of tie-points (that will be used on the bundle block adjustment);
- Survey at two different flying heights – in regards to the focal length determination;
- Also, the accuracy of self-calibration is related to the depth variations (terrain).
Take a look at these articles below:
Bolkas D., 2019. Assessment of gcp number and separation distance for small uas surveys with and without gnss-ppk positioning, Journal of Surveying Engineering, 145(3):04019007. https://doi.org/10.1061/(ASCE)SU.1943-5428.0000283
Przybilla H.J., M. Gerke, I. Dikhoff, and Y. Ghassoun, 2019. Investigations on the geometric quality of cameras for UAV applications using the high-precision UAV test field Zollern colliery, International Archives of the Photogrammetry, Remote Sensing & Spatial Information Sciences XLII-2/W13, 2019 ISPRS Geospatial Week 2019, 10–14 June 2019, Enschede, The Netherlands, pp. 531-538. https://doi.org/10.5194/isprs-archives-XLII-2-W13-531-2019
Oniga, V.E., A.I. Breaban, N. Pfeifer, and C. Chirila, 2020. Determining the suitable number of ground control points for uas images georeferencing by varying number and spatial distribution, Remote Sensing, 12(5):876. https://doi.org/10.3390/rs12050876
Hi,
First, you should use GCPs in the created test area, then you can optimize calibration parameters and save it for forwarding projects. If you also capture oblique images, it will improve the results.
Bests.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Computer Graphics and Animation
- 3D