Abstract In this paper, a systematic approach adopting sparse least-squares SVMs (LS-SVMs) is proposed to automatically detect fire using vision-based systems with fast speed and good performance. Within this framework, the features are first extracted from input images using wavelet analysis. The LS-SVM is then trained on the obtained dataset with global support vectors (GSVs) selected by a fast subset selection method, in the end of which the classifier parameters can be directly calculated rather than updated during the training process, leading to a significant saving of computing time. This sparse classifier only depends on the GSVs rather than all the patterns, which helps to reduce the complexity of the classifier and improve the generalization performance. Detection results on real fire images show the effectiveness and efficiency of the proposed approach.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    A Systematic Fire Detection Approach Based on Sparse Least-Squares SVMs

    Zhang, J. / Li, K. / Zhao, W. et al. | British Library Conference Proceedings | 2014



    Discriminative Weighted Sparse Partial Least Squares for Human Detection

    Li, Qiming / Yan, Yan / Wang, Hanzi | IEEE | 2016