在处理拉曼雷达回波信号过程中,采用小波去噪法进行算法反演前的数据预处理,针对小波去噪的核心 问题-小波基的选择和阈值的设定提出了自适应分层阈值软门限去噪的方法,并选择了不同的小波基组 合软硬阈值对去噪结果进行了对比。经实验数据验证,相对传统的滑动平均滤波,纯粹的细节和抑制全局 阈值法,自适应分层阈值软门限在获取相同信号、保持相同能量的前提下能更好地恢复信号;在有云情况 下信号的尖峰结构也没有变化,有效地抑制了噪声,提高了细节识别度和反演精度。

During the signal processing of the Raman backscattered lidar, the wavelet transform is used to reduce the noise. The setting of the threshold is the key factor in the transformation. A new method, the self-adaptive hierarchical denosing with a soft threshold is presented. Comparison of different wavelets and thresholds used in the denosing procedure are also described. Compared with the usual average method, such as detail depression method and global threshold method, the self-adaptive hierarchical denosing method can increase the SNR while with maintaining the character of the peak signal. And even when the clouds exist, the signal still can be identified without much change. The detail recognition and inverting precision are thus improved.