We have been seeing a constant vertical shift of about 2 feet between the point cloud and GCP’s and check shots on our larger projects. We are importing a .PRJ file to set up the Coordinate System output, and in this particular project it is a modified state plane system (Colorado South 0503). Everything looks great in the Quality Report. When I bring the point cloud into Trimble Business Center and process a Surface from the data, I run a surface to points report and that is where I see the constant shift. The Point Cloud/Surface is 2 feet lower than the GCP’s/Check shots. I’m not sure if it is coincidental, but the ortho images are also showing a horizontal scaling issue of about 2ft, where the image is aligned with the GCP’s well at the lower left of the image, but as you move to the upper right of the image it progressively gets off. These orthos are about a mile wide and the 2 feet of error is seen in the upper right. In AutoCAD, if I adjust the scale from the lower left of the “Raster Image” I can get it to fit perfectly all around, leading us to believe there is a scaling issue going on somewhere. Any insight into these issues would be appreciated.
In that case the problem was related to USfoot or Feet. The coordinate system was defined in USFoot but the GCPs were in International feet. Everything fit within Pix4D but when exporting the ortho into Google Earth, there was a shift.
Would you share the prj file that you have imported into Pix4D? How are the GCPs given, Feet or USFoot?
Thank you very much.
We are aware of the USfeet/International Feet issue with Pix4d. It sure would be nice to be able to set the coordinate system output to US Feet with an arbitrary system… We have been using .prj’s to get around this but have still found complications, for instance if the .prj is in a truncated State Plane system, the software thinks the coordinates are too far away from what it was expecting and gives error messages.
We have produced several more large point clouds and ortho’s in the last week, using the same .prj and are still seeing the 2ft. constant vertical shift in the point clouds. The orthos have been inconsistent in how far they are off though. We do not believe it is an international feet shift because they have all been in the correct spot in the lower left of the image and then gradually get worse as you move to the upper right, looks like a scaling issue. This particular jobsite is in a modified state plane system, where the grid projection is in the middle of the project and the projection scale factor is 1.0003538572. Here are screenshots of the GCP/MTP Manager and the .prj we are using:
Interesting that I am seeing similar problems… I have an area that is 2-3 square miles. 8 ground control points. I use very accurate dual frequency 1cm positions for the camera. We see an almost constant 42-45 centimeter vertical shift. When I mark the 8 control points as GCP’s everything is good. if we mark the 8 points as check points then we get this shift. We reoptimize the camera and turn off internal camera optimization OR we just run the standard camera setting with internal optimization turned on…the results are almost exactly the same. we get subtle differences but essentially we see a 44 cm vertical shift. If we remove the shift the results are within 10 cms vertically (3.5 cm pixel data) . We know that it is not any issues with our setup. as we confirm the elevations with the GPS unit before and after the mission.
Thank you for your feedback.
It is possible to get the output in feet or in USFoot in Pix4D, indeed, when you import the prj file that you shared, your ouput will be given in USFoot and your GCPs should be imported in USFoot. How are your GCPs given? In USfoot or International Feet?
I think your problem is different and it has to do with the camera calibration. When you use GCPs, the software is able to compute a good calibration and everything fits. However, when they are used as check points, it is not possible to compute a good calibration and the, the whole project is shifted.
Have you messed with the Geoid Height settings? Instead of using egm96?