A4 - Problem 3, Task B

Moderator: Computer Vision

robert.n
Nerd
Nerd
Beiträge: 673
Registriert: 29. Sep 2008 19:17

A4 - Problem 3, Task B

Beitrag von robert.n »

"Use the scheme from the 3rd assignment to nd the putative matches. Only consider the Harris points."

What does that mean? Not to use the Harris-Laplace approach, but instead find corner points using harris IP detection (with only one scale)?

// EDIT: I'm currently using harris-laplace.

robert.n
Nerd
Nerd
Beiträge: 673
Registriert: 29. Sep 2008 19:17

Re: A4 - Problem 3, Task B

Beitrag von robert.n »

These are the putative matches I get using the scheme from the 3rd assignment:
Bild

Total number of matches: 115

I'm using kmeans with K=3 to determine the number of inliers by counting the number of points assigned to the biggest cluster (any better ideas?). Using that approach I've determined the number of inliers to be 89 (77.4%).

//EDIT: The number of iterations thus needed is 34, if you want to randomly draw 8 inliers with a probability of at least 0.99.

Have you got similar numbers?

qgao
Moderator
Moderator
Beiträge: 33
Registriert: 7. Apr 2009 11:38

Re: A4 - Problem 3, Task B

Beitrag von qgao »

robert.n hat geschrieben:"Use the scheme from the 3rd assignment to nd the putative matches. Only consider the Harris points."

What does that mean? Not to use the Harris-Laplace approach, but instead find corner points using harris IP detection (with only one scale)?

// EDIT: I'm currently using harris-laplace.

Do NOT use Harris-Laplace, but only Harris point detection (one scale).

Antworten

Zurück zu „Computer Vision“