Improving motion segmentation with combined classifiers

Abstract

Foreground/background segmentation is an active research area for moving object analysis. Many applications in machine vision depend on high quality and robust extraction of moving objects. Established and popular methods are mixture modelling and a threshold based technique (Horprasert et al., 2000). To find a better motion classifier, a new technique is developed here, a modified Unary classifier approach that uses the bases of SVM theory. As neither the mixture modelling nor the Unary approach had implicit shadow detection, this is achieved by including colour invariant colour models. The threshold based technique has the ability to detect shadow but with the consequences of mislabelling part of the foreground. The shadow detection criterion was improved by adding a statistical constraint to the shadow detection process. In order to further extend the performance, we formed different classifiers by combining base classifiers with a Bayesian approach. The observed performance advantages are associated with the fusion of operators with complementary properties. Tests on outdoor and indoor sequences confirm the efficacy of this approach. The new algorithms can successfully identify and remove shadows and highlights with improved moving-object segmentation. A particular advantage of our evaluation is that it is the first approach that compares foreground/background labelling with results obtained from labelling by broadcast techniques, comparing a computer vision technique with an established baseline.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata