6533b82afe1ef96bd128c137

RESEARCH PRODUCT

Accurate detection and characterization of corner points using circular statistics and fuzzy clustering

Jesús V. AlbertJ. DomingoM. E. DíazF.j. FerriGuillermo Ayala

subject

Fuzzy clusteringbusiness.industryComputer scienceContrast (statistics)Pattern recognitionFuzzy logicFeature (computer vision)Pattern recognition (psychology)Computer visionNoise (video)Enhanced Data Rates for GSM EvolutionArtificial intelligencebusinessStatistic

description

Accurate detection and characterization of corner points in grey level images is considered as a pattern recognition problem. The method considers circular statistic tests to detect 2D features. A fuzzy clustering algorithm is applied to the edge orientations near the prospective corners to detect and classify them. The method is based on formulating hypotheses about the distribution of these orientations around an edge, corner or other 2-D feature. The method may provide accurate estimates of the direction of the edges that converge in a corner, along with their confidence intervals. Experimental results show the method to be robust enough against noise and contrast changes. Fuzzy membership improves the results of the algorithm and both versions (crisp and fuzzy) give better results than other previously proposed corner detectors.

yearjournalcountryeditionlanguage
1998-01-01
https://doi.org/10.1007/bfb0033293