10 likes | 90 Views
Eliminating False Positives During Corner Finding by Merging Similar Segments. 3: Output resulting corners. Aaron Wolin , Brandon Paulson Advisor Tracy Hammond. Introduction.
E N D
Eliminating False Positives During Corner Finding by Merging Similar Segments 3: Output resulting corners Aaron Wolin, Brandon Paulson Advisor Tracy Hammond Introduction Sketch recognition is the field devoted to understanding digital ink. Free-sketch recognition allows users to draw within a domain however they desire, and the recognition of the ink is left to the computer. Some recognition systems, such as LADDER, use ink object primitives as building blocks for more complex shapes. Strokes that encapsulate multiple primitives must then be broken down into their smaller parts. 2: Merge Smallest Segments Abstract Our new corner finder, MergeCF, utilizes existing primitive recognizers to accurately split longer strokes into their corresponding primitive lines and arcs. The algorithm first oversegments the stroke by finding corner candidates based on speed and curvature, and then merges unnecessary segmentations together. Primitives PaleoSketch 1 Arc Line Line Line Line • Rank by length • Test each pair using a primitive recognizer • Merge segments with low fit errors 8 2 1 4 5 10 7 3 6 9 10 11 • Initial corners found by: • > 1.00 × average stroke curvature • < 0.90 × average pen speed 1: Initial Candidates Corners found by Curvature Speed Curvature & Speed Texas A&M University