Re: [Info-gama] Gama Adjustment with Constrained/Free? with Autonomous GPS points

[Ales Cepek]

Hi Jim,

I have tried to adjust your sample data (free network) and not
surprisingly almost all distances were rejected due to huge absolute
terms. I changed the tolerance for absolute terms from 1 meter to 10
meters (from my geodetic point of view 1 meter is equivalent to
infinity but 10 meters is quite reasonable for your application).

With new tolerance no distances were rejected and adjusted
coordinates are following (and the results are surprisingly good!):

Adjusted coordinates
********************

  i        point    approximate  correction  adjusted    std.dev conf.i.
====================== value ====== [m] ====== value ========== [mm] ===

               2
  2            X *  476648.60000   1.22176  476649.82176    26.4    55.2
  3            Y * 5368157.00000   1.07884 5368158.07884    20.0    42.0

               3
 10            X *  476674.80000   2.30500  476677.10500    18.9    39.5
 11            Y * 5368116.70000   0.73033 5368117.43033    26.0    54.3

               4
  8            X *  476715.30000  -3.13487  476712.16513    27.2    56.9
  9            Y * 5368129.80000   1.00819 5368130.80819    21.4    44.9

               5
  6            X *  476685.00000  -0.01583  476684.98417    15.7    32.9
  7            Y * 5368148.50000  -1.38872 5368147.11128    18.1    37.8

              1a
  4            X *  476685.35706  -0.37606  476684.98101    17.8    37.3
  5            Y * 5368181.44580  -1.42864 5368180.01716    42.7    89.4


With the coordinate corrections in meters discrepancies in
linearization test were up to 20cm:

     ..............
  15            4 dist.     37.57000   -94.375          -50.104
  16    4       3 dir.      0.000000   915.248 3908.777 230.404
  17            2 dir.     49.190961 -1222.010 -887.198 -94.831
  18            5 dir.     57.354567   353.016  886.204  44.121
  19           1a dir.     90.703945   -46.254 -769.574 -67.959
  20            3 dist.     37.57000   -94.375          -50.104
     ..............

Thus I replaced initial point coordinates by the adjusted and results
improved substantially (my explanation is that the first sum [pvv] is
biased and not reliable).

Adjusted coordinates
********************

  i        point    approximate  correction  adjusted    std.dev conf.i.
====================== value ====== [m] ====== value ========== [mm] ===

               2
  2            X *  476649.82176   0.01791  476649.83967    12.7    26.5
  3            Y * 5368158.07884  -0.06204 5368158.01680     9.5    19.8

               3
 10            X *  476677.10500  -0.04813  476677.05687     8.7    18.1
 11            Y * 5368117.43033   0.11762 5368117.54795    12.3    25.8

               4
  8            X *  476712.16513   0.04310  476712.20823    12.6    26.5
  9            Y * 5368130.80819  -0.04546 5368130.76273     9.8    20.6

               5
  6            X *  476684.98417   0.01455  476684.99872     7.2    15.1
  7            Y * 5368147.11128   0.00287 5368147.11415     8.6    17.9

              1a
  4            X *  476684.98101  -0.02744  476684.95357     8.3    17.4
  5            Y * 5368180.01716  -0.01298 5368180.00418    19.9    41.6


In gama-local currently 'constrained' coordinates are not improved in
iterations and I am quite sure that this is correct in typical
geodetic applications (classical free networks); but your case is
quite different and I am going to add a new parameter

                   <parameters   
                               update-constr = "yes"
                   />    <!-- implicit value = "no" -->

and Jan will include it into his Rocinante later this week.

By the way, Jan is working on a new version of Rocinante so if you are
"little confused in using the Rocinante edit GUI" things might improve
a bit ;-) ... I was confused as well when Jan showed me the first
version ;-)))

My conclusion is that you can safely use Roci for you photographic
adjustment (namely when all the updates are ready; hopefully this
week).

Ales