【摘要】 本文提出一种基于加权核函数的雷达目标识别方法。该方法既对基函数加权,又对特征加权,并利用最大后验准则估计其加权系数,然后,根据权系数具有稀疏性这一特点,只保留少量有利于分类的基函数和特征元素,从而实现了基于核甬数分类器的最优化和特征选择的最优化。对三类实测飞机目标的仿真实验结果表明:加权核函数方法的识别性能比常规核函数方法有了较明显的改善。更多还原

【Abstract】 A novel method of radar target recognition is proposed in this paper.The weight coefficients of kernel ba- sis functions and weight coefficients of feature components are estimated using maximum a posteriori method,and are heavily tailed to promote the sparsity in utilization of both kernel basis functions and features,which lead to both par- simonious feature selection and excellent classification accuracy.The results of experiments on three kinds of real plane targets show that the recognition performance of the approach proposed in this paper is obviously improved com- pared with that of general kernel based method.更多还原