PDA

View Full Version : Photogrammetry?



MikeInBaltimore
05-14-2013, 12:37 PM
Would anyone have any insight on how to utilize matched pairs of spherical's to do photogrammetric measurements from said pairs?

I'm keenly interested in learning, but no real "clue" where to start looking.

:confused1:

nick fan
05-15-2013, 09:43 AM
Would anyone have any insight on how to utilize matched pairs of spherical's to do photogrammetric measurements from said pairs?

I'm keenly interested in learning, but no real "clue" where to start looking.

:confused1:

Imagemodeler supports spherical panos.
http://www.youtube.com/watch?v=C1KWP9dESDQ

Nick

MikeInBaltimore
05-15-2013, 09:51 AM
Nick,
Thanks, I'll have to chat with some of my Engineering friends. I'm not trying to really do "modeling", as opposed to extracting actual dimensions from the imagery - more like "crime scene investigation".
But this application may work.
-Mike

Hugh
05-15-2013, 01:23 PM
Hello Mike,

As you are probably aware, the accuracy of the measured points is determined by the angle the rays from the images intersect at and if you are using just two spherical panoramas then only a "donut" in the plane perpendicular to the axis between the two panos will have any reasonable accuracy.
The diagram towards the bottom of
http://hugha.co.uk/PhotoInt/PhotoInt.htm
gives a graphic for his although it is only 2D, but I am sure that you can visualise this on 3D.
There are companies that provide a photogrammetric solution from spherical panoramas, but they also provide the hardware and are very expensive, but as this is a NN forum I will not mention their names, but there are papers and the like that deal with this subject.
http://www.isprs.org/proceedings/XXXIV/5-W16/papers/PanoWS_Dresden2004_Luhmann_a.pdf
http://cipa.icomos.org/fileadmin/template/doc/KYOTO/58-1.pdf
http://rcswww.urz.tu-dresden.de/~ds627255/instpage/paper/GIM_MaasSchneider_panoramic.pdf
Presumably you are wanting to measure 3D points, but if you wish to use your spherical panoramas to measure in a plane, such as say a floor plan, then creating a Mirror Ball or Hemisphere from your panorama with the centre at the nadir will give you an image in which all the angles from the centre to the points of interest correct.

MikeInBaltimore
05-15-2013, 02:00 PM
Hugh,
THANKS! ... and what a mouthfull. I'm thinking this may be far, far over my head ... but I will investigate!
I've talked to Bill at NN and he was truthful in telling me he has no experience with this issue. They are however beginning to represent another product, the iStar ... that I am told will in the future have this capability.
So ... I'm trying to start "wrapping my head around" this ... from a humble photographer's perspective. Or ... decide to hand this issue off to our Engineering Staff :-).
-Mike

nick fan
05-15-2013, 09:08 PM
3D modeling is just one extra step than what you needed. Basically you need to identify points in the images and use photogrammetry to solve the co-ordinates in 3D.
Here is one more estabilished solution
http://www.spheron.com/?id=107

Nick

MikeInBaltimore
05-16-2013, 03:57 AM
Nick,
THANKS!
Between your comments, and Hugh's - I'd like to say "the Light at the End of the Tunnel" is beginning to get a bit brighter to this "Engineering/3D Illiterate One".
I've read about the Spheron, and a few other units which candidly ... at a starting price of ~$40K, probably WAY out of my price range.

Thank you both, and any others who might want to chime in to help this "clueless one" ...

-Mike

Hugh
05-16-2013, 12:09 PM
Hello Nick,

Thanks for giving Mike the Spheron link.
I did not actually mention this as it is surly a competitor product to your Nodal Ninjas even though it is in a different price bracket at something like 20K.
Their solution uses a rotating camera with a very narrow exposure slit (I think perhaps only one pixel wide) but is perhaps the ultimate in HDR with something like a 24 stop range.
There is another company that is a direct competitor to Spheron and in the same price bracket using a similar solution (i.e. rotating camera), but I cannot remember its name at the moment.
I'll look it up if anyone is interested.

Best regards, Hugh.

Hugh
05-16-2013, 12:21 PM
Hello Mike,

There are other methods of measuring points for things like crime scene investigations which are definitely off topic in this forum so if you wish to contact me by email I will respond.
I am now retired so don't have any vested interest, but am happy to discuss the topic.

Regarding photogrammetry, a package that is a good photogrammetric solution using "conventional" images and stereo photographs is Photomodeler.
http://www.photomodeler.com/index.html
I know that they have talked about rectification from spherical panoramas and will probably provide a solution at some stage in the future, but they are not planning to do this at the moment.

Best regards, Hugh.

Bill Bailey
05-16-2013, 01:35 PM
Hello Mike,
As you are probably aware, the accuracy of the measured points is determined by the angle the rays from the images intersect at and if you are using just two spherical panoramas then only a "donut" in the plane perpendicular to the axis between the two panos will have any reasonable accuracy.
The diagram towards the bottom of
http://hugha.co.uk/PhotoInt/PhotoInt.htm
gives a graphic for his although it is only 2D, but I am sure that you can visualise this on 3D.
There are companies that provide a photogrammetric solution from spherical panoramas, but they also provide the hardware and are very expensive, but as this is a NN forum I will not mention their names, but there are papers and the like that deal with this subject.
http://www.isprs.org/proceedings/XXXIV/5-W16/papers/PanoWS_Dresden2004_Luhmann_a.pdf
http://cipa.icomos.org/fileadmin/template/doc/KYOTO/58-1.pdf
http://rcswww.urz.tu-dresden.de/~ds627255/instpage/paper/GIM_MaasSchneider_panoramic.pdf
Presumably you are wanting to measure 3D points, but if you wish to use your spherical panoramas to measure in a plane, such as say a floor plan, then creating a Mirror Ball or Hemisphere from your panorama with the centre at the nadir will give you an image in which all the angles from the centre to the points of interest correct.

Hi Hugh,

This is good feedback. Not only have we had other Nodal Ninja users asking us about this but also iStar users as well. The applications are broad and those in the know certainly have expanded skill sets. For those that are not familiar with photogrammetry the break down of the word itself explains it:
photo = light
gram = drawing
metry = measurement

The applications are very broad - forensics (big market), architectural, surveying, archaeology, military, and the like. Here is an interesting article on photogrammetry versus lidar:
https://www.sparpointgroup.com/blogPost.aspx?id=4036

It's all interesting stuff.

Bill

Hugh
05-16-2013, 02:50 PM
Thanks Bill,

I have been involved with photogrammetry for more decades than I care to rember and am always interested in methods of using photographs for measurement so am happy to help if I can.

With respect to the iSTAR, I know that Neil and Cameron plan to develope software for measurement using the panoramic image and now that the camera is finally in the market place after all the setbacks they have encountered that perhaps we will soon see something in this direction.

Personally I would really like to see a solution for getting measurement data from spherical panoramas and think that if there is sufficient interest that it may be possible to persaude the guys at Photomodeler to look at this sooner rather than later.
Another possibility is Elcovision 10 from St Margarethen in Switzerland who's authors may consider providing this functionality - I am waiting to hear their comments.

Best regards, Hugh.

Hugh
05-17-2013, 11:10 AM
Update on previous message:

I have now heard from Elcovision that they currently support the Spheron panoramas but do not generic spherical panoramas and that although they will probably develop in this direction at some stage in the future they are not planning to in the immediate future.

I would suggest that it may be worth people contacting Mark @ Photomodeler and/or Harald @ Elcovision directly if they see a need for this functionality as my experience is that if feedback comes from a variety of sources it has more impact than if it is channeled through a single source.

Hugh
06-05-2013, 05:47 AM
Hi Hugh,

... Not only have we had other Nodal Ninja users asking us about this but also iStar users as well. The applications are broad and those in the know certainly have expanded skill sets. ...

Bill

As there appears to be some interest in this subject, I thought it would be a good idea to put some information about it on my web site:
http://hugha.co.uk/PhotoInt/SphericalPanos.htm
Having done this I am really surprised that the developers of photogrammetric software have not included the ability to measure points from Spherical Panoramas because the mathematics is so simple compared with that needed to extract measurements from photographs, particularly if Equirectangular projections are used where simple arithmetic only is required to extract the Horizontal and Vertical angles to the points of interest.

Unfortunately I have not written any software for over a decade and never ventured into manipulating digital images. I have however written a program to automatically computed co-ordinates from theodolite angular measurements from different locations taking into account and weighting for the angles of cut and rejecting observations not within a specified tolerance, so do have a fair idea of what is involved.
What is required is a program to extract the pixel co-ordinates from Equirectangular projections, allowing zooming in and panning around the images, then using the overall pixel dimension of the Equirectangular projections to calculate the Horizontal and Vertical angles by simple arithmetic as shown on:
http://hugha.co.uk/PhotoInt/SphericalPanos.htm
then use trigonometry to compute the co-ordinates.

Anybody out there willing to take on this challenge?
I am willing to make my source code for computing the intersections available, or providing a pseudo-code program.


Hugh,
THANKS! ... and what a mouthful. I'm thinking this may be far, far over my head ... but I will investigate!
I've talked to Bill at NN and he was truthful in telling me he has no experience with this issue. They are however beginning to represent another product, the iStar ... that I am told will in the future have this capability.
So ... I'm trying to start "wrapping my head around" this ... from a humble photographer's perspective. Or ... decide to hand this issue off to our Engineering Staff :-).
-Mike

Yes Mike, I quite agree that the papers I linked to complicate the subject, in my view unnecessarily, but I think that if you go back to your Engineers with the information I have just published that they may take a new view as getting the Horizontal and Vertical angles from an Equirectangular projection
is really very straight forward and this data is exactly the same as an Engineer/Surveyor would get from observing with a theodolite on site for theodolite intersection with the added benefit that little time is need to make the panoramas, especially with an iSTAR, compared with the much greater time need for the theodolite observations, even with a modern instrument and that unlike with theodolite measurements only you have the imagery to refer back to in the office.

MikeInBaltimore
06-05-2013, 05:54 AM
Gentlemen, and I do mean everyone ... I can't thank you all enough for your responses!
So very much to wrap my head around ... I'm trying but LOL, I have a hard time spelling "Engineer" ... :-)

MikeInBaltimore
06-05-2013, 04:21 PM
Hugh - I can't help but chuckle to myself a bit.
After reading thru most of your published material ... I actually think it's a bit 'scary' to think there are folks out there this 'smart'.

With deep respect,
-Mike

Hugh
08-13-2013, 08:20 AM
For those interested in this topic, and its potential, I have add some more information to my web site page:
http://hugha.co.uk/PhotoInt/SphericalPanos.htm
including some related trigonometry and notes on the vertical separation of the panoramas as well as horizontal.

MikeInBaltimore
08-13-2013, 08:43 AM
Hugh ... all I can say is "WOW!" ... and I need to talk to my Engineering friends again.

Hugh
08-17-2013, 03:31 AM
Thanks Mark,

I am planning to put a post on the LinkedIn Photogrammetry group in the hope that we can find some talented people willing to take up the challenge of creating the software tools to make quantitative data extraction from spherical panoramas an automated process.