Generic line scan camera's ground to image iteration is very inefficient

[jessemapel]

The generic line scan camera uses Newton's method to calculate sample and line. Looking at some of the literature on this problem, that looks like a possibly poor method.

See:
https://www.asprs.org/wp-content/uploads/pers/1993journal/may/1993_may_655-661.pdf
https://www.asprs.org/wp-content/uploads/pers/2001journal/april/2001_apr_449-452.pdf
https://www.asprs.org/wp-content/uploads/pers/2009journal/september/2009_sep_1059-1067.pdf

We should at the very least switch to solving for line (down-track) and then sample (across-track). ISIS uses a bisecting window search method, that we should consider implementing.

[jessemapel]

Another good reference about pushbrooms in general
http://users.rsise.anu.edu.au/hartley/public_html/Papers/pushbroom/pushbroom.pdf

[jessemapel]

More references
http://www.isprs.org/proceedings/XXXVII/congress/3b_pdf/14.pdf

[jessemapel]

CCPS search is probably not going to work because it requires consistent integration times and viewing angles.

[jessemapel]

We should start with the bisecting window search.

From Wang et al:

The first step of this method is to choose an image interval between the first and the last scanline as the initial search window. Then, it iteratively halves the search window by comparing the three pairs of image point coordinates calculated through colinearity equations, respectively, by employing the EO parameters of the first, the middle, and the last scanline of the search window. Finally, a sequential search is carried out in the final bisecting window within a threshold size

This is basically what ISIS3 does

[jessemapel]

Another more detailed resource for bisection search, see page 42

https://earth.esa.int/documents/700255/1929094/N77D7-PV02-US-5-IQC-GC-ATBD-ACS-0102-2.1/582f28ab-707c-4640-997c-8073d16d224f