全变差正则化法凸优化平滑γ能谱

Convex Optimization Based Total Variation Regularization Smoothing for γ Spectra

在残差代价最小化函数基础上加入全变差正则化惩罚函数项,通过凸优化技术实现了γ能谱的平滑,并提出了由最优权衡曲线确立残差和全变差正则化权衡因子方法。与滑动平均、SavitzkyGolay、小波滤波等方法的对比结果表明,该方法参数设置简单便捷,在全道域内即保证了平滑后能谱峰形的相似,又获得了更好的平滑效果。

The minimization of the total variation regularization method for γspectra smoothing is studied and solved by the convex optimization technique. Through the trade- off curve the method for the determination of the optimal trade- off factor, which compromising the cost function and the penalty function namely the total variation regularization, is introduced. The comparison between the traditional methods including moving averaging, Savitzky-Golay in addition to wavelet denoising and the method proposed shows that the latter is more effective as a whole in such peak zones as the lower peak disturbed by a higher peak nearby, doublet multiplets and the big statistical fluctuation peak. Besides, the parameter of the proposed method is easier to be determined.