DensifyPointCloud

[Yangyp29]

Why use the point cloud generated by the DensifyPointCloud, which appears to be truncated by a section at a distance, and the objects further away are not reconstructed? What parameters can be improved by adjusting?(As shown in the picture below, the left side of the picture is a reconstructed distant object)

[cdcseacave]

not sure what you mean, OpenMVS reconstructs by default everything that is seen in the input images. Check that DensifyPointCloud does not use a ROI or some kind of bounding box

[Yangyp29]

As shown in the image above, we input images taken by 12 cameras, in which the tower and the tower in the distance are visible in the images.
We check the Densify options in DensifyPointCloud. Use default values for all options except the following options:
-i
out_openMVS/scene2.mvs
-o
scene2_dense.mvs
--roi-border
0
--estimate-roi
0
--crop-to-roi
false

However, the resulting point cloud is shown below, with a front view on the left and a side view on the right(It looks like distant objects have not been reconstructed or that distant points are all concentrated at the same depth). Why does this happen? How to adjust?

[Yangyp29]

not sure what you mean, OpenMVS reconstructs by default everything that is seen in the input images. Check that DensifyPointCloud does not use a ROI or some kind of bounding box

As shown in the image above, we input images taken by 12 cameras, in which the tower and the tower in the distance are visible in the images.
We check the Densify options in DensifyPointCloud. Use default values for all options except the following options:
-i
out_openMVS/scene2.mvs
-o
scene2_dense.mvs
--roi-border
0
--estimate-roi
0
--crop-to-roi
false

However, the resulting point cloud is shown below, with a front view on the left and a side view on the right(It looks like distant objects have not been reconstructed or that distant points are all concentrated at the same depth). Why does this happen? How to adjust?
As shown in the image above, we input images taken by 12 cameras, in which the tower and the tower in the distance are visible in the images.
We check the Densify options in DensifyPointCloud. Use default values for all options except the following options:
-i
out_openMVS/scene2.mvs
-o
scene2_dense.mvs
--roi-border
0
--estimate-roi
0
--crop-to-roi
false

However, the resulting point cloud is shown below, with a front view on the left and a side view on the right(It looks like distant objects have not been reconstructed or that distant points are all concentrated at the same depth). Why does this happen? How to adjust?

[cdcseacave]

wow, this is not good image acquisition, no surprise you get such a bad reconstruction!!

first of all it seems the images are taken from inside a plane, flying passing by the scene; if you want to reconstruct the tower (or anything in this world) you need to cover it with images from all possible angles, the more the better, but the technology is nowhere reconstructing them from 1-2 images from the same view point, not yet!