I'm a new user to PI, and so far I'm quite impressed with a great deal it has to offer. However, I'm having a problem trying to assemble a complex mosaic. I've spent the last year gathering data around the North Celestial Pole. What I did was to capture 24 panels with a 50mm lens and DSLR with each panel having Polaris close to one side and extending to approximately +67 degrees on the other (each panel is about 25 degrees by 16 degrees with the long edge running north-south). Each panel was centered on an hour of R.A., so that I basically have 24 pie wedges radiating from the NCP (of course, each "wedge" is actually a rectangle). There is plenty of overlap between each of the "wedges", with every wedge overlapping near the NCP and a smaller amount of overlap at the +67 degree side. When I'm done assembling the mosaic, I basically have a circle with a radius of about 23 degrees centered on the NCP.
I'm attempting to assemble the mosaic using the same process that Steve Allan demonstrates in his
"Creating a Seamless Super Mosaic" video (thank you Steve for doing this video!). The problem I'm having is that I can assemble anywhere from about 8 to 13 or so of the panels using Star Alignment's Register/Union Mosaic mode (for creating the initial "map"). However, once I get to a certain point, the registration fails with an "Error: Unable to find an initial set of putative star pair matches".
What I've tried so far is:
1) I created a distortion map for the 50mm lens (this helped quite a bit with registration, as the initial panels assembled much more quickly).
2) I've tried both 2-D Surface Spline and Projective Transformation models. I have not changed the Spline smoothness setting from the default of 0.25.
3) After I assemble panels one and two, I'm checking the "Undistorted reference" box. My theory here is that once I have a two panel or more mosaic as the reference image, the distortion map no longer matches that image (it does match the next panel that I'm adding).
4) I've checked the "Distortion correction" box. I've mostly been using it with the distortion residual setting at .200 pixels and 50 iterations, but I've also tried it with the maximum .500 residual and 100 iterations (after I've had a panel fail).
5) I've tried starting the mosaic with different panels (i.e., one time I started with 0 hrs, another time with 3h, 6h, etc.). It doesn't seem to be a problem with a particular panel, as it will fail with different panels (and succeed with that same panel if I start with a different panel). Typically, it'll fail after I have around half of the circle completed.
6) I tried assembling roughly a third of the circle (0h to 9h), separately assembling the second third of the circle (9h to 17h) and then trying to align those two overlapping mosaics. It fails when trying to register the two mosaics together (even though they have plenty of overlap at the center of the mosaic as well as between about the 7h panel to the 12h.
7) I was able to register the three 1/3 circles in Registar, but I'd prefer to use PI for this step (my concern that Registar may warp the panels and then PI will have problems with the next step - creating the panels that match the full size mosaic). But, this at least confirmed for me that there isn't some reason the panels can't fit together.
Is there a different method or tool I should be trying to use? Would it help if I used the Image Solver to solve each of the panels before assembling? If so, should I have it solve each of the mosaics as I add panels to it before adding the next panel or just wait until it fails before solving the mosaic?
Any other suggestions?
Scott
http://www.astronomersdoitinthedark.com/
