摘要
由于雷达目标及其所处环境的复杂性,导致目标之间的关系往往是非线性的,因此,基于核方法的模式识别方法被广泛应用于雷达目标识别中。在对模糊核C-均值聚类算法深入研究的基础上,提出一种基于模糊核C-均值聚类的高分辨距离像识别算法。该算法针对特征提取后一维距离像数据的特点,采用组合核函数以降低由于数据属性数值过大造成的权重过大对识别效果的影响;同时,算法可以在训练过程中通过有效性函数自适应地确定最佳聚类数目。仿真实验结果表明,基于组合核函数的识别算法同基于传统的高斯核的算法都能有效识别雷达目标,但前者具有更高的目标识别率。
Because of the complexity of radar targets and the environment,the relations between targets are usually nonlinear.Consequently,the pattern recognition algorithm based on kernel method is applied widely in radar target recognition.In this paper,one HRRP recognition method based on fuzzy kernel clustering was proposed.A new kernel function was applied according to the properties of the range profiles.This new kernel function can reduce the impact of over-weighs caused by the oversized attribute of the data.Further more,this method can determine the best clustering number using the validity function in training progress.Experiments show that the recognition method based on this new kernel function and the algorithm based on Gauss kernel function can both recognize the radar targets,but the first one has higher identification ratio.
出处
《电光与控制》
北大核心
2010年第5期42-45,共4页
Electronics Optics & Control
基金
国家自然科学基金资助(60572138)
重点实验室基金资助(9140C8001020902)
