Year: 2016
Type: conference-paper
Abstract: An efficient recognition framework requires both \ngood feature representation and effective classification methods. \nThis paper proposes such a framework based on a spatial Scale \nInvariant Feature Transform (SIFT) combined with a logistic \nregression classifier. The performance of the proposed framework \nis compared to that of state-of-the-art methods based on the \nHistogram of Orientation Gradients, SIFT features, Support \nVector Machine and K-Nearest Neighbours classifiers. By testing \nwith the largest vehicle logo data-set, it is shown that the proposed \nframework can achieve a classification accuracy of 99.93%, \nthe best among all studied methods. Moreover, the proposed \nframework shows robustness when noise is added in both training \nand testing images.
Institutions: University of Sheffield, University of Birmingham
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2407901846