6533b82afe1ef96bd128cc37

RESEARCH PRODUCT

New Error Measures to Evaluate Features on Three-Dimensional Scenes

Fabio BellaviaDomenico Tegolo

subject

Ground truthFeature Detector Feature Descriptor Overlap Error Epipolar Geometry Feature Matching and ComparisonSettore INF/01 - InformaticaFeature (computer vision)business.industryEpipolar geometryFeature descriptorPattern recognitionArtificial intelligencebusinessFeature matchingMathematics

description

In this paper new error measures to evaluate image features in three-dimensional scenes are proposed and reviewed. The proposed error measures are designed to take into account feature shapes, and ground truth data can be easily estimated. As other approaches, they are not error-free and a quantitative evaluation is given according to the number of wrong matches and mismatches in order to assess their validity

yearjournalcountryeditionlanguage
2011-01-01
https://doi.org/10.1007/978-3-642-24085-0_54