Quantitative Comparison of Feature Matchers Implemented in OpenCV3
Zoltán Pusztai, Levente Hajder · SZTAKI Publication Repository (Hungarian Academy of Sciences) · 2016
The latest V3.0 version of the popular Open Computer Vision (OpenCV) framework has just been released in the middle of 2015.The aim of this paper is to compare the feature trackers implemented in the framework.OpenCV contains both feature detector, descriptor and matcher algorithms, all possible combinations of those are tried.For the comparison, a structured-light scanner with a turntable was used in order to generate very accurate ground truth (GT) tracking data.The tested algorithm on tracking data of four rotating objects are compared.The results is quantitatively evaluated as the matched coordinates can be compared to the GT values.