Hough Detection/CS question

[Kuciwalker]

I know next to nothing about statistics because at my high school it's either calculus or statistics, and I picked calc (no regrets there). I don't have any idea what the mathematical way to identify an outlier would be.

abs(X[i] - mean) > C*sqrt(sum((X - mean(X))^2)) where C is a suitable constant

note this is only one approach and may not work in your case depending on the distribution, if so try some of these: http://en.wikipedia.org/wiki/Outlier...fying_outliers

also, read the AP statistics guide, it should take you a few hours and you'll learn everything you need for the exam

[Kuciwalker]

but ultimately I prefer that other kid's method, of just finding the most likely one. presumably your algorithm returns a likelihood?

[Hauldren Collider]

It doesn't return a likelihood. And yeah, his method is pretty good. But I also want to try it my way . I talked with my teacher today and this is really easy with standard deviations. Here's the idea:

If something is roughly circular, we can assume it'll follow some sort of normal distribution. A point on the circle is more likely to be close to the approximate circle than further away. So you sort the list of radii. The list of radii for two concentric circles size 2 and 3 will be something like 2,2.01,2.1,2.101,...3,3.02,3.05,3.1,3.3... and so on. Then what you do is iterate across the list, taking an average as you go along, and at each element you calculate the standard deviation (I know how to do that) and see if the element you're currently at exceeds the standard deviation. If it does, you take your current average and store it as one of your circles. Then you start a fresh average...rinse lather repeat. If something doesn't have sufficient points (some arbitrary threshold) you throw it away. This eliminates noise. Noise will generally be at the end of the list, so you might be able to truncate or something.

I haven't implemented this yet, but if you like I'll post it somewhere so you can see it once I get it working. After that, I'll try squares or triangles

[Kuciwalker]

There are better clustering algorithms. For this you probably want to iterate k-means clustering until increasing k provides less than some threshold increase in fidelity.

[MRT144]

The best! (according to US News and World Reports, anyway)

https://www.tjhsst.edu

Our website kinda sucks now. I'm one of the students that runs the school's computer system. Believe me, we had another proposal for the design, this one wasn't our choice

Seriously, there aren't any network admins there? WTF!

Also, the caption for the physics class photo is needlessly pretentious. Pondering physics? Cmon.

[Hauldren Collider]

Seriously, there aren't any network admins there? WTF!

There are network admins, for example, the windows network is run by the staff. Actually our faculty admins are amazing, best I've ever seen in a public school. However we have two UNIX labs (one solaris + sun ray, one linux workstation) that are run entirely by students. Students also run the website and a fair bit of the network (a couple of the Cisco switches, for instance). There's around 10 student sysadmins and we do pretty much everything not windows.

Also, the caption for the physics class photo is needlessly pretentious. Pondering physics? Cmon.

STFU. There is absolutely no description of Dr. Dell that could possibly describe how awesome he is. I do not exaggerate when I say that he's probably more brilliant than most college professors. I think he used to be a professor but he quit because he hated it.

[MRT144]

There are network admins, for example, the windows network is run by the staff. Actually our faculty admins are amazing, best I've ever seen in a public school. However we have two UNIX labs (one solaris + sun ray, one linux workstation) that are run entirely by students. Students also run the website and a fair bit of the network (a couple of the Cisco switches, for instance). There's around 10 student sysadmins and we do pretty much everything not windows.


STFU. There is absolutely no description of Dr. Dell that could possibly describe how awesome he is. I do not exaggerate when I say that he's probably more brilliant than most college professors. I think he used to be a professor but he quit because he hated it.

Awesome!