I am trying to do a panorama model of an environment, namely a street, basically a long, straight plane for the street and then elementary geometry for the sides of the street.  Initially I tried to do it with a Stitcher panorama, as the manual suggests this is possible; after reading this forum, I realize that it is not.  However, if I save the panorama out of Stitcher as a cubic map, I can get two images to do a very good calibration.  However, I assumed that if one image was calibrated and I added points to it that were also in a third image, that the third image would be calibrated.  In other words, the two camera locations are projecting a cube from Stitcher; the Y rotation of the cube is slightly offset, so the left image from location 1 has some of the front image from location 2 and some of the left image from location 1.  So, I take the two left images and they calibrate well; I take the front image and put 8 new points in it, and also put those points in the image from the other location, but the new image does not calibrate.  It's like it doesn't even try, even if I try to force it.  Am I trying to the impossible?  Is it possible to take a long street - like a block - and model it in ImageModeler?

Hello Jon,
 
Generally the following situation is not recommended :
 
3 images I1, I2, I3
I1 and I2 have markers in common
I2 and I3 have markers in common
I1 and I3 have no marker in common
 
But, here are some advice that could help calibrating in this situation:
- Constrain the Camera parameters
    - distortion is known and its value is 0 (as Stitcher did the undistortion)  
    - the focal of cubic render is known !
         focal = filmbackWidth / 2.0 in this case   (as Width==Height)
- Add "simple"shots that helps covering your envirronment
Create a different camera Device for those shots.
 
Hope it helps,
 
Stef

Ok, I guess I'm having a bit of a challenge getting my head wrapped around what can and can't be done with IM... As I said, I want to model about a block of a street (and really more, I'm doing the block to test to see whether it can be done); there is no position where I can put the camera and see enough that it will share markers without at least one trio of images as you describe I1-I2 have common markers, I2-I3 have common markers, I1-I3 have no common markers.  In fact, I have to take a picture about every 20 feet to really get a good selection of common points between pictures.  How would I go about doing this?

Hello Jon,
 
  Yes you need to walk and take pictures along the road.
 
I would do something like the attached drawing.
-> take pictures on both side of the road
-> take some pictures backwards (opposite to the walking direction)
 
Hope it helps,
 
Stef

As an IM neewbie Iam experiencing the same problem.  In order to model a complete building or series of buildings, as you are, you will always encounter the problem of unrelated images - of the type you are experiencing.  In many ways I can see potential for this application in my practice and I realise that the secrete is in the calibration but it all seems a little too random to me.  I just can't set up the scene that the second camera object calibrates with the first? Is there any one who can explain it to me, differently please?