An adaptive-weight regularization method for multi-classifier fusion decision

Abstract

The difficulties of aircraft type recognition methods are introduced and the necessity of multi-classifier fusion decision method is discussed. The some kinds of invariants: Hu moments, Affine moments, Zernike moments, Wavelet moments, are used for constructing four SVM classifiers. Based on the above four classifiers, an adaptive-weight regularization method is proposed for improving aircraft type classification performance. Experiments are shown that, the recognition rate by the proposed method in this paper is better than any classifier of the above four classifiers, the fixed-weight multi-classifier fusion method and the majority multi-classifier fusion method.