摘要
针对基于高分辨距离像(HRRP)多类特征联合舰船目标识别的问题,提出了一种基于Fisher判决率加权的修正最近邻模糊分类器。在对舰船目标的HRRP特性进行分析的基础上提取船长、离散性、对称性、中心距等稳定特征,结合各类特征的稳定性和可分性,设计一种能让不同特征充分发挥优势作用的修正最近邻模糊分类器。该分类器用Fisher判决率对特征差隶属度进行加权修正;通过10类军民船目标的实测数据验证,表明基于Fisher判决率加权的修正最近邻模糊分类器在舰船目标识别领域具有很好的实际应用前景。
Modified nearest neighbor fuzzy classification algorithm based on Fisher discriminantweighting is proposed for ship target recognition using multi-features of high resolution range profile(HRRP). Firstly, ship length, dispersant, symmetry and central moments features are extracted. Then,modified nearest neighbor fuzzy classification algorithm is designed for different feature to contribute theirpredominance because the stability and separability of each feature are different. The membership degree ofeach feature is modified by Fisher discriminant weighting differently. The result of experiment with theactual measured data of 10 ship targets indicates that the proposed classifier is very useful in ship targetclassification.
出处
《舰船科学技术》
2010年第2期68-72,82,共6页
Ship Science and Technology
关键词
高分辨距离像
Fisher判决率
修正最近邻模糊分类器
中心矩
high resolution range profile ( HRRP)
Fisher discriminant
modified nearest neighborfuzzy classification algorithm
central moments