基于多不变量和多分类器融合的飞机识别

Aircraft recognition based on multiple classifier fusion with multiple invariants

分析了飞机图像自动识别方法的难点,指出在飞机图像识别中采用多分类器融合的必要性。提出利用飞机图像的4种不变量特征:仿射矩,Zernike矩,小波矩,SIFT特征点梯度模值,结合支持向量机组成4种分类器,采用自适应权重投票法进行多分类器融合,以提高飞机机型识别率。仿真实验表明,上述4种图像不变量特征构造的多分类器,经过自适应权重投票法融合判别后,飞机机型识别率明显优于单一种类不变量特征构造的同类分类器,同时优于固定权重投票法、多数投票法的多分类器。

The difficulties of aircraft image automatic recognition method are analyzed,and the advantages of using multiple classifier fusion for aircraft recognition are pointed out.Then,four kinds of invariants,which include affine moment,Zernike moment,wavelet moment,gradient module of SIFT feature descriptor,are combined with support vector machine to produce four kinds of classifiers.An adaptive weighted voting method is adopted to carry out multiple classifier fusion for improving aircraft recognition rate.Simulation experiment results show that using the proposed adaptive weighted voting method of multiple classifiers,the aircraft recognition rate is superior to the recognition rates using the classifiers constructed with single invariants,and also superior to those using the multiple classifiers constructed with fixed weighted voting method and majority voting method.