S.B., AsmaD., Abdelhamid2021-01-182021-01-182020978-172812190-1DOI: 10.1109/M2GARSS47143.2020.9105140https://www.scopus.com/record/display.uri?eid=2-s2.0-85086715489&origin=SingleRecordEmailAlert&dgcid=raven_sc_affil_en_us_email&txGid=cc2c6ecdf7c879938a36b62b38e690c2https://dspace.univ-boumerdes.dz/handle/123456789/6169This paper introduces a novel method for the classification of very high resolution, multispectral, remote sensing images. We combine the advantages of both pixel-based and object-based classification techniques. First, the pixels contained in the image are grouped into different batches, called segments, using the algorithm of superpixels. Then the superpixels are merged into more significant objects using one distance metrics among a variety. Finally, the resulting image is classified by the Support Vector Machines classifier. The performance of the proposed approach is compared to the classical spectral-based classification. Using overall accuracy and average accuracy, the results obtained on a high resolution multispectral Boumerdes image reveal the efficiency of the proposed methodenImage ClassificationObject-BasedRemote SensingSuperpixelsArticle