I have a problem getting the model in correct location with ground control points. I have four surveyed control points with XYZ-coordinates. I mark the points on several photos and then align images and reconstruct the model. The model gets correct scale and orientation from ground control points.
BUT:
The problem is to get the model on correct location. It seems like RC always put the model in center of the coordinate system (X=0 Y=0 Z=0). Is there any solution to calculate the model on correct location?
I have a problem getting the model in correct location with ground control points. I have four surveyed control points with XYZ-coordinates. I mark the points on several photos and then align images and reconstruct the model. The model gets correct scale and orientation from ground control points.
= OK
BUT:
The problem is to get the model on correct location. It seems like RC always put the model in center of the coordinate system (X=0 Y=0 Z=0). Is there any solution to calculate the model on correct location?
in what sw you importing the model ?
go to WORKFLOW/SETTINGS, look for COORDINATE SYSTEM, OUTPUT system is local or ?
what coordinate system you use ?
can post some screens ?
Maybe I explained poorly. I’m not importing model into the RC. I’m just doing basic workflow (adding photos, setting ground control points, align photos and reconstruction)
Here is my workflow with screencaptures:
I load 250 photos to RC
The ground control points are black&white targets that have known coordinates (coordinates surveyed with Leica total station).I set each target in several photos (seven photos for each target in this case). After setting the targets I change point type from tiepoint to ground control point and type coordinates for each. Coordinate system: local:1 - Euclidean.
Then I align images and do reconstruction. In 3d view the model is in correct orientation (long sides of the building is parallel with X-axis and short sides of the building paraller with Y-axis) The problem is that, the model is moved in the center (0,0,0).
Interestingly in “component 0” the ground control points are in correct location:
Shortly my question is:
How can I keep the model stay in correct location and not moving to the center?
Dear Jane,
the component 1 is the correct one. The thing is that the application automatically center the object so that it can be visible in the 3D view. However it is not a mistake it is intended to be so. Next you can compute model and then an ortho projection. Then you will see the correct coordinates for the mouse possition in the left down corner of the ortho projection 2D view. Furthermore if you set “Preserve Coordinates” to true during export of the model then the exported model will contain correct coordinates of the model vertices as well.
Dear Jane,
the component 1 is the correct one. The thing is that the application automatically center the object so that it can be visible in the 3D view. However it is not a mistake it is intended to be so. Next you can compute model and then an ortho projection. Then you will see the correct coordinates for the mouse possition in the left down corner of the ortho projection 2D view. Furthermore if you set “Preserve Coordinates” to true during export of the model then the exported model will contain correct coordinates of the model vertices as well.
Yes it works! Thanks!
Does anyone know what is the best method to get the best location accuracy? Is it better to set ground control points in every photos where the black&white targets are visible? Or is it better to set ground control points only some photos and let the software do the calculations. By decreasing the number of manually tagged GCPs the human error is smaller and it also saves time.
I’m used to make coordinate based registration on AutoDesk ReCap. In Recap 4 manually tagged photos for one GCP is enough to make a good result
Does anyone know what is the best method to get the best location accuracy? Is it better to set ground control points in every photos where the black&white targets are visible? Or is it better to set ground control points only some photos and let the software do the calculations. By decreasing the number of manually tagged GCPs the human error is smaller and it also saves time.
it depend on project, but as you say adding additional CPs (GCPs) on images is time consuming and error prone by human, and in short say 10 imgs with properly set GCP should be OK. As adding more, the error is decreased just “insignificantly”
But from own long year work with it im tend to use 4-6 projections as under 4 get still “significant” error.
I did some pointcloud comparison between Reality Capture and Recap. Same 250 photos and same four GCPs used in both. Reality Capture has a lot more detail! :shock:
Dear Jane,
the component 1 is the correct one. The thing is that the application automatically center the object so that it can be visible in the 3D view. However it is not a mistake it is intended to be so. Next you can compute model and then an ortho projection. Then you will see the correct coordinates for the mouse possition in the left down corner of the ortho projection 2D view. Furthermore if you set “Preserve Coordinates” to true during export of the model then the exported model will contain correct coordinates of the model vertices as well.
How can I import a model from another program that has vertices in the same coordinate system as the GCPs, so that it is placed properly?
Hi Wishgranter,
I tried that and it comes in at the wrong place. I also exported a mesh from RC with Preserve coordinates and imported it back into RC and it is in the wrong place.
Dear Biil,
you must export your model (modelName.obj) with info file (set to true the “General Settings -> Export info file” in the “Export Model” dialog and it will create modelName.rcinfo). This file must be available also side by side with the modified model you are about to import. If you are importing modifiedModelName.obj then just copy the modelName.rcinfo to modifiedModelName.rcinfo … If the application finds the rcinfo then it will use it to place the model to correct local coordinate system.
It worked for the model exported from RC. I tried to import a PS mesh by changing the .rcinfo file to the same name as the PS mesh file name, but it didn’t work for that one. Both meshes are referenced using the same control points.
It worked for the model exported from RC. I tried to import a PS mesh by changing the .rcinfo file to the same name as the PS mesh file name, but it didn’t work for that one. Both meshes are referenced using the same control points.
My mistake, when I changed the name of .rcInfo file to the the name of the PS file I just used .rcInfo as the extension. I went back and added .obj to the extension and it worked. e.g. PS_name.obj.rcInfo
Hello,
I have a project and i need to know if there is any pattern on how to mark targets gcps, materials etc.
The flight altitude is about 50m and i want to find the accuracy of gcps.
