I have a project where some cameras are not well aligned and I can solve this by uncalibrating them, adding a few tie points and recalibrating, which places them in a new position, often correct.
Now, this project has a lot of straight lines, which could be modeled with polylines.
Unfortunately, these polylines have only two points (start/end), and we always need to “fully qualify” the points (I mean, mark the point in a few images).
How could we add a lot of valid tiepoints along these polylines without having to mark each one in several images? With such feature, image alignment and calibration would be piece of cake.
Hi Daniel. This seems like something that would be easily done in GIS with the orthomosaic. If you don’t have access to and ESRI products you can use QGIS (Free alternative) to digitize the features in your raster outputs generated with Pix4D.
But I’d want this “to generate” a quality orthomosaic in Pix4D, for aligning the cameras properly.
I’m not sure I understand. The orthomosaic is not used to calibrate the external or internal camera parameters. MTPs in the pointcloud can be used to improve the results of step 1 but once you are processing in step 3 there is no camera calibration taking place. What do you mean when you say “quality”? Is there a specific issue you are encountering or are you speaking generally? The best way to ensure good quality results in step 3 is to ensure that step 1 has processed correctly and that there are no uncalibrated cameras, the internal optimization is low, and that the project had processed as a single block. All of these factors are reported at the beginning of the report.
I would like to generate tie points at regular intervals over a polyline, in order to have a good step 1 with good calibration (1), in order to have a better point cloud and a better orthomosaic.
The idea of this feature is:
Instead of marking every single tie point manually
Mark only the start and end points of a straight line
Populate this line with fully qualified tie points (meaning they’re indeed marked in 3 or more photos, showing green in the cloud, not orange)
PS: the system lets us add extra points to a polyline, but these points are not fully defined (they need manual action of being marked on images)
(1) because my project is full of straight bands with perfectly smooth surfaces and the program doesn’t match the points correctly for these bands. I cannot physically mark these bands to take new photos, these objects should not be touched on the field.
Ok, thanks for clarifying. I think there are better ways to ensure good camera calibration. Rather than generating MTPs like this you should be flying with enough overlap with a camera suitable for mapping and that usually is enough. I’m not sure why you need to generate so many MTPs if you are using a verified workflow.
When you say the program “doesn’t match tie points correctly” this may be due to several reasons. You may want to consider opening a support request and submitting a quality report for review along with a description of the problem.
Well, this is a suggestion for a new feature that would certainly help us (and probably many other users).
We face this kind of mismatch quite often (long reflective/smooth surfaces/bands). While going back to the field present a lot of additional problems, travelling again, getting a lot of troublesome authorizations for restricted areas, more airplane tickets, etc., solving this via software would certainly be a great help.
We do solve lots of these by using surfaces and marking them to be part of building the orthomosaics, but sometimes a simple polyline, if taken into account in camera calibration would solve half or our problems.
Thanks for explaining Daniel. I will pass your suggestions on to the devs so that they can consider them for future versions of Mapper. Thanks for sharing your feedback.