I am doing a project measuring coastal erosion over time. I am using photogrammetry of the same location over difference to evaluate changes, using the cloud-mesh distance tool. I am having two problems, can you help?
1) Meshes of the same location, imported with identical coordinate systems, are projected in CloudCompare at different scales and apparent locations, even though the meshes were created from high resolution geotagged air photos 3 cm per pixel.
2) After "Matching Bounding Box Centers" and "Matching Scales", I align the meshes by "Picking Equivalent Point Pairs". After picking 15-20 tie points spread across the meshes and hitting the align button, I find that a majority of the tie points do not completely align. Using the "Fine Registration ICP" tool fails to improve the alignment accuracy. In consequence, the resulting computed distance layer produces false change. The output has areas that indicate distance change but in reality are caused by the misalignment of meshes. This systematic misalignment makes it impossible to distinguish real areas of erosion from alignment errors.
Has anyone else had these problems? Does anyone have any insights? I am happy to talk more about these issues.
Need advice on aligning meshes and calculating distance changes
Re: Need advice on aligning meshes and calculating distance changes
1) I believe that it's inherent to the photogrammetry reconstruction process (the lack of reproducible scale). You should normally used Ground Control Points or other means to scale the meshes properly
2) With ICP, are you changing the 'Overlap' parameter? This is mandatory if both meshes don't fully overlap.
2) With ICP, are you changing the 'Overlap' parameter? This is mandatory if both meshes don't fully overlap.
Daniel, CloudCompare admin
Re: Need advice on aligning meshes and calculating distance changes
Our meshes were made with geotagged images and they were fully georectified using ground control points.
We did adjust the final overlap parameter, testing values between 40% and 95%. These changes did not increase the alignment accuracy of the meshes.
In spite of all this, CloudCompare continues to project our meshes at different scales and locations, and I cannot get the meshes to align as described above.
I am a new student user and wondering if there are other steps or tricks that I might be unaware of to better align the 2.5D meshes.
We did adjust the final overlap parameter, testing values between 40% and 95%. These changes did not increase the alignment accuracy of the meshes.
In spite of all this, CloudCompare continues to project our meshes at different scales and locations, and I cannot get the meshes to align as described above.
I am a new student user and wondering if there are other steps or tricks that I might be unaware of to better align the 2.5D meshes.
Re: Need advice on aligning meshes and calculating distance changes
Well, if you don't check the 'rescale' option, then the alignment tools will never change the scale.
And for the coordinates, make sure you choose the reference entity wisely. And also mind that when a Global Shift is applied at loading time, then it's only restored at export time. And some mesh formats don't manage large coordinates properly (not sure what's your exact process).
And for the coordinates, make sure you choose the reference entity wisely. And also mind that when a Global Shift is applied at loading time, then it's only restored at export time. And some mesh formats don't manage large coordinates properly (not sure what's your exact process).
Daniel, CloudCompare admin