Timeline for Ellipse-detection algorithm
Current License: CC BY-SA 3.0
15 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 1, 2017 at 12:30 | comment | added | Gykonik | And I added an second edit to my post :) | |
| Feb 1, 2017 at 12:02 | comment | added | Gykonik |
Okay! Two questions, in your optimized code bSquared return negative numbers. I think, that you typed something wrong there. Also I wanted to ask, how you would make the "average"-Thing on which I'm currently failing. Maybe you could also add this to your optimized code! And again you don't know how greatful I am to you :)
|
|
| Jan 31, 2017 at 22:48 | comment | added | Peter Taylor |
I think that the authors of the paper are using an array of ints for their accumulator, indexed by rounded-off semi-minor axis. But I think it makes more sense to store the exact calculated semi-minor axis (because the cluster may not be centered on an integer) and also to store the value of k (as an optimisation for removing the points). (I'm also starting to think that it would be better to take points i, j, and the ones from the winning cluster and put them into an ellipse best fit process rather than continuing to assume that points i and j define the major axis exactly).
|
|
| Jan 31, 2017 at 20:17 | comment | added | Gykonik |
okay, I thought the same! Okay, I, I think, that I understand everything except for "Increment the accumulator for this length of minor axis". What should be stored in accumulator?
|
|
| Jan 31, 2017 at 19:44 | comment | added | Peter Taylor | @Gykonik, I'm not convinced that there's much that can usefully be cached. To understand step 10, look again at step 8: "Increment the accumulator for this length of minor axis by 1". It seems that they're quantising the minor axis (and table 1 bears that out). So then in step 10 "Find the maximum element in accumulator array" means "Of the candidates for minor axis, find the one which maximises the number of points on the ellipse". But if no minor axis gives you at least e.g. 5 points, there isn't an ellipse at all and the agreement of e.g. 4 points is just random noise. | |
| Jan 31, 2017 at 19:23 | comment | added | Gykonik | And could you Explain step 10 in the paper? "If the vote is greater than the reqired least number for assumed ellipse, one ellipse is detected" I don't understand the meaing of this... | |
| Jan 31, 2017 at 18:16 | comment | added | Gykonik | And what Do you think about caching? :) | |
| Jan 31, 2017 at 18:16 | comment | added | Gykonik | Perfect, i'll look through this tomorrow! :) | |
| Jan 31, 2017 at 18:00 | comment | added | Peter Taylor | @Gykonik, have you seen my first edit (second revision), which expanded on the "inconsistent use of continue"? I've just made a third edit which adds some links to data structures and expands slightly on the other points you mention. | |
| Jan 31, 2017 at 17:57 | history | edited | Peter Taylor | CC BY-SA 3.0 |
added 348 characters in body
|
| Jan 31, 2017 at 17:45 | comment | added | Gykonik | Thanks, I have a few more questions: What Do you mean with geometric lookup Data structure? What's wrong with the average thing? What do you mean by "inconsistent use of continue"? Could you maybe explain? And could you explain step 10 in the paper:"If the vote is greater than the reqired least number for assumed ellipse, one ellipse is detected " because I don't unterstand it :/ (maybe you could add the points in your answer :)) | |
| Jan 31, 2017 at 16:07 | history | edited | Peter Taylor | CC BY-SA 3.0 |
added 3173 characters in body
|
| Jan 31, 2017 at 14:37 | history | edited | Peter Taylor | CC BY-SA 3.0 |
Expand on a couple of points which weren't clear enough
|
| Jan 31, 2017 at 13:14 | review | Suggested edits | |||
| Jan 31, 2017 at 13:58 | |||||
| Jan 31, 2017 at 11:46 | history | answered | Peter Taylor | CC BY-SA 3.0 |