Citation:
Wolin, Aaron, Brian Eoff, and Tracy Hammond. "ShortStraw: A Simple and Effective Corner Finder for Polylines." SBM. 2008.
Publication Link: http://www.researchgate.net/profile/Tracy_Hammond/publication/220772398_ShortStraw_A_Simple_and_Effective_Corner_Finder_for_Polylines/links/0deec529f76e58d523000000.pdf
Summary:
This paper presents an alternative algorithm for corner detection, where the earlier work has been done by Sezgin, Yu and Cai. The main phases of this algorithm are Resampling, Bottom up Phase and Top Down Phase. The overall result was a massive improvement over the previous two techniques, in both corner accuracy as well as all-or-nothing accuracy. The biggest drawback of this algorithm however, is its failure to detect obtuse corners.
Discussion:
The main takeaways here are the 3 phases of the Shortstraw algorithm:
1. Resampling
The first step is to obtain a resampled representation by using an interspacing distance.
2. Bottom-up
A straw for a point is computed as the distance between the endpoints of a window containing the point.(window size is generally 3 on either side of the point). The points having short straw length are determined as corners, using a threshold of of (median * 0.95)
3. Top-Down
(i) First, it checks to see if every segment between two consecutive corners passes a line test (chord distance/path distance)
(ii) Removing false negatives: If any two corners fail the line test, it is assumed that there must a missing corner, approximately between them, and the threshold is relaxed to find it.
(ii)Removing false poistives: Checking for co-linearity among triplets of consecutive corners.
Wolin, Aaron, Brian Eoff, and Tracy Hammond. "ShortStraw: A Simple and Effective Corner Finder for Polylines." SBM. 2008.
Publication Link: http://www.researchgate.net/profile/Tracy_Hammond/publication/220772398_ShortStraw_A_Simple_and_Effective_Corner_Finder_for_Polylines/links/0deec529f76e58d523000000.pdf
Summary:
This paper presents an alternative algorithm for corner detection, where the earlier work has been done by Sezgin, Yu and Cai. The main phases of this algorithm are Resampling, Bottom up Phase and Top Down Phase. The overall result was a massive improvement over the previous two techniques, in both corner accuracy as well as all-or-nothing accuracy. The biggest drawback of this algorithm however, is its failure to detect obtuse corners.
Discussion:
The main takeaways here are the 3 phases of the Shortstraw algorithm:
1. Resampling
The first step is to obtain a resampled representation by using an interspacing distance.
2. Bottom-up
A straw for a point is computed as the distance between the endpoints of a window containing the point.(window size is generally 3 on either side of the point). The points having short straw length are determined as corners, using a threshold of of (median * 0.95)
3. Top-Down
(i) First, it checks to see if every segment between two consecutive corners passes a line test (chord distance/path distance)
(ii) Removing false negatives: If any two corners fail the line test, it is assumed that there must a missing corner, approximately between them, and the threshold is relaxed to find it.
(ii)Removing false poistives: Checking for co-linearity among triplets of consecutive corners.
No comments:
Post a Comment