Map difficulties

[paul womack]

Following my recent success (thanks to those who helped)
de-tessalating a map, I've attempted to revisit
a previous example.

This one is comprised of (all)
the Boundary Commission Maps, 1885
from http://www.londonancestor.com/maps/maps.htm.

Each ward was shown on a single page of a book.
In order to get as much detail in as possible
they used different scales as appropriate,
and also rotated the maps, so that North
isn't always "up".

(I was having great difficulty using these maps
until I noticed that little fact... :-)

I have successfully detessalated these
maps into a large composite by
my usual dirty trick of declaring
them to have a very small field of view
(AKA long lens), and then optimising
in the usual "giga pan" style for
Y P R. Due to the changing scale
I also gave each map a separate
lens and optimised for HFOV too.

This all worked nicely.

However, I now feel that I should
do this panaorama "properly"
by using X,Y.

I gave it a try, but I couldn't
make it work.

Questions:

* Do I still need a separate lens for each map
do handle the scaling, or will Z motion deal
with that (is there a write up of X,Y,Z model?
the values seem quite odd).

* since some of the maps are rotated
do I need to optimise for R as I did
in my gigapan approach?

BugBear

[Terry Duell]

Hello Paul,

On Thu, 06 Feb 2014 20:10:25 +1100, paul womack <pwo...@papermule.co.uk>
wrote:

[snip]

> Questions:
>
> * Do I still need a separate lens for each map
> do handle the scaling, or will Z motion deal
> with that (is there a write up of X,Y,Z model?
> the values seem quite odd).
>
> * since some of the maps are rotated
> do I need to optimise for R as I did
> in my gigapan approach?
>

Give each a new lens and optimise r,X,Y,Z.

Cheers,
--
Regards,
Terry Duell

[paul womack]

Terry Duell wrote:

>
> Give each a new lens and optimise r,X,Y,Z.

OK, I've tried that; the optimiser essentially failed.
Came up with very small HFOV (0.9 !!) and
set all the X and Y to zero,
and had a range or Z's from -1.2 to 0.
(what is the unit of X Y Z).

So - no dice.

Since the problem is well defined by
the pto, I attach it here, in it's "old" form.

BugBear

[paul womack]

paul womack wrote:
> Terry Duell wrote:
>
>>
>> Give each a new lens and optimise r,X,Y,Z.
>
> OK, I've tried that; the optimiser essentially failed.
> Came up with very small HFOV (0.9 !!)

OK - rule for hugin. Never try to optimise ALL the FOV's,
always leave one (perhaps the one on the anchor...) set!

I have now succeeded in making an XYZ model
from the per-optimised YPR model, but have
not succeeded in getting an XYZ model to optimise
from a fully reset parameter set.

BugBear

[Terry Duell]

Hello Paul,

On Fri, 07 Feb 2014 20:42:14 +1100, paul womack <pwo...@papermule.co.uk>
wrote:

I had a look at this project, and managed to get it to optimise, but not
really sure if the result is a sane one.
Attached screenshot of preview and the resulting .pto.

[paul womack]

Terry Duell wrote:
>>
>> OK, I've tried that; the optimiser essentially failed.
>> Came up with very small HFOV (0.9 !!) and
>> set all the X and Y to zero,
>> and had a range or Z's from -1.2 to 0.
>> (what is the unit of X Y Z).

>

> I had a look at this project, and managed to get it to optimise, but not really sure if the result is a sane one.
> Attached screenshot of preview and the resulting .pto.

That works nicely (I rotated it so that North is up), by altering the R
of the anchor a little at a time, re-optmising after each change.

So - follow-up questions

1) How did you do it?!
2) what are the units of X Y Z (it's clearly not pixels).
I'm also most unclear on why, given the nature of this
data (and the use of multiple lenses), Z is varying at all.

And yet; X and Y only vary between -0.1 and 0.1, but Z
runs from -0.5 to 1.2

I'm clearly misunderstanding X Y Z, which is probably
why I'm not making good use of the modelling
they provide.

BugBear

[Terry Duell]

Hello Paul,

On Mon, 10 Feb 2014 21:03:06 +1100, paul womack <pwo...@papermule.co.uk>
wrote:

[snip]

> That works nicely (I rotated it so that North is up), by altering the R
> of the anchor a little at a time, re-optmising after each change.
>
> So - follow-up questions
>
> 1) How did you do it?!

I didn't make any notes, but my recollection is that I reset yaw and pitch
of all images, then set v,r,X,Y,Z to be optimised for all except the
position anchor.

> 2) what are the units of X Y Z (it's clearly not pixels).
> I'm also most unclear on why, given the nature of this
> data (and the use of multiple lenses), Z is varying at all.

Don't know the answer to that.

[paul womack]

Terry Duell wrote:
> Hello Paul,
>
> On Mon, 10 Feb 2014 21:03:06 +1100, paul womack <pwo...@papermule.co.uk> wrote:
>
> [snip]
>> That works nicely (I rotated it so that North is up), by altering the R
>> of the anchor a little at a time, re-optmising after each change.
>>
>> So - follow-up questions
>>
>> 1) How did you do it?!
>
> I didn't make any notes, but my recollection is that I reset yaw and pitch of all images, then set v,r,X,Y,Z to be optimised for all except the position anchor.

I just tried that (quite carefully) and got an optimiser report of:

avg: 157
std: 572
max: 2819

Horrible!

Can you rack your memory, or go from the start again keeping notes?

BugBear

[Terry Duell]

Hello Paul,

On Wed, 12 Feb 2014 22:42:28 +1100, paul womack <pwo...@papermule.co.uk>
wrote:

[snip]

>>> 1) How did you do it?!
>>
>> I didn't make any notes, but my recollection is that I reset yaw and
>> pitch of all images, then set v,r,X,Y,Z to be optimised for all except
>> the position anchor.
>
> I just tried that (quite carefully) and got an optimiser report of:
>
> avg: 157
> std: 572
> max: 2819
>
> Horrible!
>
> Can you rack your memory, or go from the start again keeping notes?
>

Oh dear, sorry if my recollections were wrong.
I'll have another look at it, and make notes, and hopefully Murphy's Law
won't get in the way.
I will have time to try again in a couple of hours.

[Terry Duell]

Hello Paul,

On Wed, 12 Feb 2014 22:42:28 +1100, paul womack <pwo...@papermule.co.uk>
wrote:


>

> Can you rack your memory, or go from the start again keeping notes?
>

OK, attached london-4.pto, mean error 3.0, max 40.1.

The process I went through is as follows, in Expert interface, Optimiser
tab;
1. 'unselect all' Yaw (y) (right click on column header)
2. 'unselect all' Pitch (p)
3. 'select all' X, Y , Z (image #2 is automatically not selected as it is
position anchor)
4. deselect hfov(v) for lens #2 (position anchor)
5. Optimise

...and Bob's your Aunt's main squeeze :-)

Hope that helps.