Underside of a bridge

[adrian pope]

I have taken a number of photos looking up at the underside of a bridge to try to produce a rectangular image of the cracks. 10m by 8m photographed from 4m below.

I loaded all the image into a tiny free program, with no controls, and it produced an almost perfect stitch, with perfect vertical lines (it is a series of parallel beams) but with major barreling top and bottom. The photos appear to be suitable for stitching.

I have tried for 20 hours now to use Hugin to get a better result but I am missing something major. I have 50% overlaps, with 30+ control points between each but have gotten nowhere. Even when I tell it that there are verticals, and add in verticals for the full height of a photo, it still bends the output and has poor registration between images.

I finally used Xara (like coreldraw) to individually perspective correct each of three images (as a trial) so that the "Vertical" joints were vertical, and the "Horizontal" shutter marks (also with control points assigned)  were horizontal but it seemed distorted the images badly and gave an awful colour output.

I have tried using mosaic as well as photosphere and every projection I can find but  all to no avail.

Should Hugin be able to do what I want or am I wasting my time

Many thanks

[Bruno Postle]

This should be possible as long as the surface you are surveying is flat.

The only output projection that will display straight lines in the 3d scene as straight lines in the 2d stitched output is 'rectilinear'.

Are all the photos taken from the same position? In which case this is a 'normal' panorama, though you will need to set horizontal and vertical control points to remove any remaining perspective 'keystone' distortion.

If the photos are each taken from different positions, then this needs to be stitched as a 'mosaic'. The technique depends on the photos, but usually trying to optimise all parameters in one go doesn't work so well. So you may need to optimise x, y & z positions first, then bring in r, p & y rotations, and finally add the lens parameters.

--
Bruno

[adrian pope]

Thank you for the reassurance. The underside is flat (except for the joints which step in about an inch) but with a small precamber so that the middle is about 60mm high over 10mm.

The photos are from two locations and I have tried both together (panorama and mosaic) and separately but the prog does not overlay well. The beams have a number of perpendicular shutter marks and I have called up a lot of these as horizontal, but something else seems to be over-riding this. The auto vertical finding gave short lengths and I added a number of full length ones to help orientate the photos. I have attached a much reduced image for interest :-)

I have not seen where to optimise x y and z, despite trying expert mode.

I will have a hunt through the automatic pairs of points to see if they have caused a problem. I seem to recall an error of 7 (what ?) being shown during stitching (7 pixel average error) but if someone would like a suggestion for the prog then an option of highlighting those matched pairs with the highest error would be useful

Thank you for the help.

A

[adrian pope]

For someone who prides himself on his intelligence, and his spacial awareness, I can be a complete idiot sometimes. Apart from the errors in the matching, I was trying to get a rectilinear image off of a spherical projection, but my camera was offset by about 10% of the image (I was lying in the road behind bollards with traffic running a metre away from my head and could not move further across). It is not possible to correct for the field of view that I want since the normal from me to the plane will hit off centre. When I resize the field of view to include more of the sphere outside of what I want then it gets closer.

It is a shame that the controls for yaw etc have to be types in and cannot be controlled by up and down arrows but I am getting there.

Many thanks

On Monday, 12 July 2021 at 08:30:37 UTC+1 bruno...@gmail.com wrote:

[Bruno Postle]

Hi Adrian, you should be ok with a slight camber, though this may be
visible as a 'kink' in the final result.

Select Geometric -> Custom Parameters in the Photos tab, this will
enable X, Y & Z parameters in the Optimise tab.

There is a pop-up 'Control Point table' where you can find
control-points with high error distances, the numbers are pixel
distances in the final panorama output.

--
Bruno

On Mon, 12 Jul 2021 at 09:24, adrian pope wrote:
>
> Thank you for the reassurance. The underside is flat (except for the joints which step in about an inch) but with a small precamber so that the middle is about 60mm high over 10mm.
> The photos are from two locations and I have tried both together (panorama and mosaic) and separately but the prog does not overlay well. The beams have a number of perpendicular shutter marks and I have called up a lot of these as horizontal, but something else seems to be over-riding this. The auto vertical finding gave short lengths and I added a number of full length ones to help orientate the photos. I have attached a much reduced image for interest :-)
> I have not seen where to optimise x y and z, despite trying expert mode.
> I will have a hunt through the automatic pairs of points to see if they have caused a problem. I seem to recall an error of 7 (what ?) being shown during stitching (7 pixel average error) but if someone would like a suggestion for the prog then an option of highlighting those matched pairs with the highest error would be useful

--
Bruno