基于多分类支持向量机的船舶桨叶数识别研究

Ship propeller blade number classification based on multi-class support vector machine

分析了目前常用的支持向量机多分类方法以及存在的不足,本文提出了一种混合纠错输出编码的多分类支持向量机改进算法,并应用于利用船舶目标辐射噪声DEMON谱进行船舶桨叶数分类的实验。理论分析与实验结果表明,该改进算法编码明确、具备纠错能力,是一种有效的多分类支持向量机方法,在船舶桨叶数识别中,其分类性能优于一对余、一对一及最小输出编码支持向量机等多分类方法,可适用于船舶桨叶数的分类识别。

The paper first analyzes the common multi-class support vector machine （SVM） classification algorithms and points out the disadvantages of these algorithms, and then presents an advanced multi-class SVM classification algorithm based on mix error correcting out code （MECOC）. Experiment of ship propeller blade number classification which is based on DEMON spectrum of ship target radiated noise has been done by using this algorithm. Theoretical and experimental results show that the proposed algorithm with clear code and error correction ability is an efficient multi-class SV1VI classification algorithm. In the ship propeller blade number classification experiment, the classification performance of this algorithm is better than one-versus-all, one-versus-one and minimum output coding （MOC） multi-class SVM classification algorithm, which is suitable for ship propeller blade number classification problem.