a little problem i am working on

Today i would like to talk about a little problem i have that most people don't see as a problem...but i sure do. I will start by pointing out the pretty picture to the left. It is a little study area i have designated B5. This is B for Buchanan Bay and 5 for a reason that no doubt made perfect sense years ago but somehow i never really wrote down why. This is is a "sample area" of the East Antarctic land mass that i like to use for testing different algorithms for how well they do a particular job and for how much processing time/power gets use up.

Trust me the bigger the data set the more inclined you are to worry about program efficiency! Plus, you are less likely to piss off other computer users (especially when your using something as valuable as a super computer run by some one else...and don't pay for it...but they do.....OLD story)

At the beginning of my research i use to pick a couple of images and use them for all my testing but the trouble was that i never had enough variability in the images i used and later on all sorts of unforeseen stuff would happen and takes AGES to sort out...and was quite dull. If you consider what i do you can only imagine how dull YOU would find it.


One of the problems i grapple with is that i want the computer to do all the analysis of an area. The thing is i want it to be as close as possible the same area each time. The following image shows the different types of problems you can encounter. The red line is a coastline. The trick is to look at how closely and in what way the image deviates.





What i now want to do is take this little sample of all my images and work on percentage of good data to bad data....with a somewhat less clearly defined "bad data" than one might think. I will do the same for another, more (i was going to say northerly.....but that only works for MY warped version of north) rather a less longitudinal area. I will then get some sort of pattern to when bad data is likely to crop up....


The thing is that it is to do with the satellite viewing angle, problems with various algorithms and sometime instrument problems. There was also the famous 1994 problem when a winter's worth of data was not collected because no at the station (Casey) realised that the tape machine was not writing data to the tapes. Yes. They sent back an entire season of BLANK tapes. Yes. They changed the tapes over every 12 hours for months...for nothing. We all had a good laugh at that one. Oh yes.


It is important when your using any data for your research that you understand the exact nature of that data. You should never except what other people write or say about it. Find out for yourself.


Now, depending on how well this little blog goes done (and how i feel on the day) you will get a blog about a Wiener Filter (which took longer to get working that i care to admit) or another story about me (or my family, or TV or....well, you get the idea).