This paper presents a new procedure for fitting multiple geometric structures without having a priori knowledge of scale. Our method leverages on Consensus Clustering, a single-term model selection strategy relying on the principle of stability, thereby avoiding the explicit tradeoff between data fidelity (i.e., modeling error) and model complexity. In particular we tailored this model selection to the estimate of the inlier threshold of T-linkage, a fitting algorithm based on random sampling and preference analysis. A potential clustering is evaluated based on a consensus measure. The crucial inlier scale ϵ is estimated using an interval search. Experiments on synthetic and real data show that this method succeeds in finding the correct scale.

Scale estimation in multiple models fitting via consensus clustering

Magri L.;
2015-01-01

Abstract

This paper presents a new procedure for fitting multiple geometric structures without having a priori knowledge of scale. Our method leverages on Consensus Clustering, a single-term model selection strategy relying on the principle of stability, thereby avoiding the explicit tradeoff between data fidelity (i.e., modeling error) and model complexity. In particular we tailored this model selection to the estimate of the inlier threshold of T-linkage, a fitting algorithm based on random sampling and preference analysis. A potential clustering is evaluated based on a consensus measure. The crucial inlier scale ϵ is estimated using an interval search. Experiments on synthetic and real data show that this method succeeds in finding the correct scale.
2015
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
978-3-319-23116-7
978-3-319-23117-4
Multi-model fitting
Scale estimation
Segmentation
File in questo prodotto:
File Dimensione Formato  
paper.pdf

accesso aperto

: Post-Print (DRAFT o Author’s Accepted Manuscript-AAM)
Dimensione 6.25 MB
Formato Adobe PDF
6.25 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11311/1188382
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? 1
social impact