CAJ | 학술논문

提出了一种基于傅里叶-梅林变换和二叉树支持向量机相结合的舰船目标一维距离像识别方法.该方法充分利用了傅里叶-梅林变换具有的时移与尺度不变性和支持向量机在小样本分类中的优势,可以改善目标的特征稳定性,提高识别性能.针对多类舰船目标的识别,提出采用聚类分析中的均值距离来生成二叉树,将分类器分布在各个节点上,构成了多类支持向量机,减少了分类器数量和重复训练样本的数量.对4类舰船目标仿真实验的结果表明,该分类方法具有较高的识别性能、较快的识别速度.
제출료일충기우부리협-매림변환화이차수지지향량궤상결합적함선목표일유거리상식별방법.해방법충분이용료부리협-매림변환구유적시이여척도불변성화지지향량궤재소양본분류중적우세,가이개선목표적특정은정성,제고식별성능.침대다류함선목표적식별,제출채용취류분석중적균치거리래생성이차수,장분류기분포재각개절점상,구성료다류지지향량궤,감소료분류기수량화중복훈련양본적수량.대4류함선목표방진실험적결과표명,해분류방법구유교고적식별성능、교쾌적식별속도.
This paper presents a new method of ship recognition based on the Fourier-Mellin trans-form (FMT) and the binary tree support vector machine (SVM). The method makes full use of the FMT's property of shift and scaling invariance and the SVM's advantage in small-sample classification to improve the stability of target characteristic and raise the recognition performance. To solve the multi-class problems, the average class distance of clustering is used to construct binary tree. The method distributes classifiers to each node which constitutes multi-class SVM, so it can reduce the number of SVM classifiers and repetitive training samples. The experimental results on range profiles of four targets show that the method is feasible.