Work

A MKL-MKB Image Classification Algorithm Based on Multi-kernel Boosting Method

Year: 2016
Type: conference-paper
Abstract: Aiming at the low accuracy and poor applicability of traditional SVM classifiers, this paper proposes an image classification system based on MKL-MKB (multi kernel learning-multi kernel boosting). This approach firstly integrates existing feature extraction methods to extract features like wavelet, Gabor, GLCM and so on. A weak classifier is constructed by using a synthetic kernel in kernel space. We use Nystrom approximation algorithm to calculate weights of kernel matrixes of multi-kernel model. Then we make a decision level fusion of weak classifiers under Adaboost framework to impair weights of weak kernels. Finally, experiments are carried out to verify the validity and applicability of the proposed algorithm by testing on terrain remote sensing images and several UCI data sets.
Institution: Beihang University
Language: English
Open Access status: hybrid
OpenAlex ID: W2557379196
FWCI:
Cites: 16
Cited by:
Related to: 10