This is for internal testers at the moment. English Tutorial is coming soon.
You must have the basic knowledge of surveying the ground control point (GCP) in order to use this feature. In some countries or regions, license is required in surveying ground control points. Please follow the local laws and regulations. At the moment, the testing is by invitation only. If you want to get an invitation, please email us at firstname.lastname@example.org. We’ll send you one as soon as we can provide more seats for testing.
Here in this step, you can only see the camera poses with an orthophoto if you add GCPs for the first time in this project (just as the picture shown below).
Click Import on the top-left panel to upload GCP data.
But before moving on, let's first understand the functions of Export. Literally speaking, it means to export the coordinates data that you have already marked on the map in .json file. You can use this function when you want to keep a record.
After you finish marking the GCP positions in at least one photo of a Ground Control Point (which we'll talk about later), our machine will regard this point as "real" and record its position. So, when you Export the coordinate data, the .json file you export will include the locations of this GCP.
If you do the same thing for all the GCPs, then you'll have all the coordinates in the exported .json file.
There are generally two ways to import data.
The first is to import it in .txt format, while the second is .json format.
If you are importing the coordinates into this project for the first time, you can only choose to import them in .txt file.
For .json format, as explained above, it is just another option for you to import data.
Zoom in the photo and mark your GCP in the photo precisely.
Click Save. Or otherwise, the machine will not record the changes you make.
Do the same thing in 5-7 photos for each GCP so that you can triangulate an accurate 3D ground control point.Tip: There’s an Eye icon on the bottom-left of the photo. That is the toggle of a green crosshair that represents the original GCP position in the preliminary marking. Click it, and you can switch on/off the green crosshair.
Tip: It would be better to select the GCPs in the centre of your orthophoto as checkpoints, rather than the ones near the corners or edges.
It sounds a bit complicated, but actually it’s as simple as that. To understand it, just imagine that building a 3D model is like putting all the photos in their right positions corresponding to a place on the 3D model. So, in the SFM stage, the machine will predict the position of each photo. But this is just a preliminary estimation and the machine is not 100% sure about it. However, by adding GCPs in the reconstruction process, the machine now has something to refer to. Based on these GCPs, the machine can slightly adjust the positions it previously predicted.
After you marked all the GCPs in photos and started a preliminary calibration, our software would try to fit your GCPs into the camera poses (SFM) result without making dramatic changes to each camera pose. The software will then generate a Preliminary GCP Report, telling you the fitting deviation that the SFM result might have based on the GCPs as you have marked.
However, some people may wonder, since Preliminary GCP Calibration doesn’t make any adjustments to each camera pose, what’s the point of doing it? The main purpose is to help measure the accuracy of the GCP positions that the user has marked in photos. In this way, the user can identify the errors and make necessary corrections in time.
Space Error: (Unit: meter) The distance between the 3D coordinate triangulated from the GCP positions that you have marked on 2D photos, and the absolute 3D coordinate of this GCP. Please notice that the distance here is in 3D.
Ground Error: (Unit: meter) The distance between the projected point of the 3D coordinate triangulated from the GCP positions that you have marked on 2D photos, and the projected point of the absolute 3D coordinate of this GCP.
Height Error: (Unit: meter) The distance between the height of the 3D coordinate triangulated from the GCP positions that you have marked on 2D photos, and the height of the absolute 3D coordinate of this GCP.
Reprojection Error (Unit: pixel) The GCP positions that you have marked on 2D photos, will triangulate a 3D coordinate. This coordinate will then be reprojected back to each photo of this GCP. The Reprojection Error is the distance between the reprojected point on a photo and the GCP position that you have marked on this photo.
Below is a Preliminary GCP Report with no big errors:
Possible Reasons for the Errors in GCP Report:
Height Error is big:
Ground Error is big:
Reprojection Error is big:
The GCP positions in different photos of this control point are not totally identical to each other.
Ground Error is small, while Height Error is big:
When you import the GCP coordinates, you flip the X, Y coordinates.
If you can't figure out the reasons, please use the Report Problem function on the project page. Our technicians will help look into it.
In Preliminary GCP Calibration, the machine will not make any real changes to camera poses or the position of each photo. But in Complete GCP Calibration, the machine will make actual adjustments to the camera poses.
Below is an excel report for Complete GCP Calibration with no big errors.