I'm guessing this is a question of edge detection... but I might be barking up the wrong tree with that...
Have a look at the image, Sample_Transition.gif. This is about the cleanest version I can find of a process that happens about every minute or two across every day. "X" varies from 0 to 24 hours, "Y" varies from 44 to 120 (in this case). In the example, the transition from 44 to 90 is stepwise (which I'm not particularly interested in); that is, it generally rests on 44 and "if the conditions are right", it will immediately jump to 90; there are no in-between values. It's the next region that I'm more interested in...
What I'm trying to do is to detect when Y becomes somehow stable on 90 - let's call that T1. Then, Y will get to a stage of being somehow stable on 120 - let's say that's at T2. For simplicity's sake, I'll make the decision that I would like T3, the time I'm interested in, to be such that T3 = (T1+T2)/2, so I get a rough idea of the mid-point of the transition from "stable Y point 1" to "stable Y point 2".
As you can see in the periods around 06:40 and 07:40, there is still some instability and I'm guessing I might need to decide what constitutes "stable" before I can go any further... but how can I do that when the silly thing flips around so much? (a more typical picture is Busier_Activity.gif) - note that there should be another stable Y point in the period 09:30-15:00 or so).
Perhaps I need to do a line-fit of some sort, to get an "average" straight line for each of these "bouncing" regions and that would be enough (for example, see Simplified_Activity.gif)... but I'm not even sure how to reduce all this noise to something simpler...
I'd greatly appreciate any suggestions on how I might attack this...
Thanks.
NOTE: MediaFire is having update problems so the images are currently only downloadable and not directly viewable. Hopefully, this condition won't last long. --Aug 19, 2009 at 01:35 UTC
Update:Given up with MediaFire, trying something else for the image hosting, which seems to work Ok. --Aug 24, 2009 at 00:06 UTC
In reply to How to Determine Stable Y-Values... or Detect an Edge..? by ozboomer
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |