摘要
从目标识别实际问题出发,研究了将动态支持向量机应用于高分辨距离像目标识别算法。根据目标特征与待识别目标之间的距离定义惩罚函数,给每一训练样本赋一惩罚参数,体现出不同样本对待识别目标的不同贡献,并根据惩罚参数大小重新构建训练样本集。由于以某一个具体目标的识别为核心,不寻求全局性的分类面,因此具有较好的针对性和动态性。
The application of dynamic support vector to the HRRP target recognition from the perspective of practical problems is studied in this paper. The penalty function was defined based on the distance between the target feature and the target to be recog- nized, assign a penalty parameter to each training sample in order to show their different contributions to the target. Then the training sample set was reconstructed based on the penalty parameter. Since we have made a specific target as a core and do not seek global classification, the method is thought to be more effective and dynamic.
出处
《现代雷达》
CSCD
北大核心
2011年第12期43-46,50,共5页
Modern Radar
基金
国家自然科学基金(61171155)
中国航天科技集团公司航天科技创新基金(CASC200902)
关键词
动态SVM
高分辨距离像
惩罚函数
目标识别
dynamic support vector machine
HRRP
penalty function
target recognition
