CE 603 Homework 3 Bundle Block Adjustment
- retrieve all of the matlab & data files from below and
begin to process the block. you will need to inspect and possibly
edit six input data files, in addition to the 26 individual
photo files that i just received from you for homework 2.
- the six files to inspect/modify are: cam.dat, cp.dat,
delta.dat,pho.dat,phofiles.dat,sig.dat. the 26 individual
photo files are "phx_y.dat" as you sent them to me.
see the file "anno.txt" for an annotated listing of all of the
files with descriptions of what needs to be there. examples of
things you will need to add are control points, photo station
initial angles, photo filenames, etc. examples of things that
will need to be inspected and edited are: point name consistency.
- when you have all of the input prepared, run pba.m. run from
the matlab command line - its not a GUI. output should be self
explanatory (i hope). you probably need to run on the "big" version
of matlab - i'm guessing that the student version will choke
on the large matrices. (note: another reason to consider the
sparse matrix tricks!).
- each person should prepare the data independently and
run/debug the block individually. if it is determined that
one individual's data is desparately in need of "fixing" we
may ask that individual to fix it. if that happens and it is
you - please respond.
- our goal is to get a good adjustment without have to
reject too many pass points, and hopefully no control points.
good means overall sigma-x,y on the order of a few pixels.
(60-70 um). we may have to add some functions and listings
to the program. a good debug strategy is to set the
number of iterations to 1 (variable: max_iter) and look
at the residuals of only the first iteration. to be effective
the initial approximations need to be good.
- due thursday, 13-feb. i need: the listing of results,
notes on any points that had to be discarded, remeasured, renamed
etc. i.e. what did you have to do to get it to work, and what
are any lessons learned if you had to do this again (you will.).
- looking ahead, the next thing that we will do is to look at
the propagated uncertainties in the computed parameters and how
they influence intersections and stereo collection, etc. also
we do some rectification and pairwise rectification for stereo.
and we do a terrestrial, small format block using the chem-e points
as control.
- version 1 of program, data, and
documentation files (.zip)
- revised data files (.zip) no more HUGE errors, 13-feb-03
- matlab program to plot photo points (.zip)
- documents describing passpoints (.zip)
- questions you should worry about: how sensitive are we to
initial approximations? how do additional rays beyond the minimum
influence the propagated uncertainty of a ground point? what
is the benefit of multiple rays in finding observation errors?
how does the presence of blunders influence the processing? what
can we do about it? if you were in a well-oiled production shop,
how long would it take to do this job (hint, it would not be a
month). how could we evaluate the results externally as well as
internally? using our results, what is the largest scale at which
we could map and still meed map accuracy standards (1/2 mm H, 1/2
c.i. V)? would these results have been changed by scanning at
10 or 15 micrometers instead of 30? at what points could we
reliably insert automation into this process?
Back to CE603 Home Page