为了提高差分光柱像运动激光雷达(DCIM雷达)探测信噪比, 提出了一种基于集合经验模态分解(EEMD)和奇异值分解(SVD)的混合降噪法.由EEMD获得含噪信号多层模态分量, 根据各模态分量之间互相关系数的差分量确定主要噪声并予以滤除, 利用奇异值分解识别模态分量中的残余噪声并提取有用信号.利用混合降噪法EEMD-SVD和EEMD方法分别对模拟仿真信号和实测激光雷达信号进行降噪处理.结果表明, 当模拟噪声标准差在0.05~0.2之间时, 相比与未降噪直接反演的湍流廓线, EEMD-SVD方法降噪后反演的湍流廓线信噪比提高了2.718 7 dB~6.921 5 dB, 相应的EEMD方法提高了1.446 1 dB~3.366 1 dB; 两个不同时段DCIM雷达降噪前后反演廓线与探空廓线的对比发现, EEMD-SVD和EEMD两种方法降噪后反演廓线较之于未降噪的反演廓线, 信噪比最大提高了2.526 5 dB和2.155 6 dB.EEMD-SVD的降噪效果优于EEMD, 能够更有效地识别和滤除噪声, 较大地提高了原始信号的信噪比, 获得更准确的大气湍流廓线反演结果.

In order to enhance the Signal-to-Noise Ratio (SNR) of Differential Column Image Motion lidar (DCIM lidar) detetion, a hybid denoising method which combines Ensemble Empirical Mode Decomposition (EEMD) and singular value decomposition(SVD) is proposed.The multilayer mode components are obtained from EEMD firstly. The difference of cross-correlation coefficients among these mode components is then utilized to determine the main noises which should be removed. The residual noises contained in mode components are identified by SVD and then the useful signal is extracted. Both the EEMD-SVD and EEMD methods are used to denoise the simulation signals and measured DCIM lidar signals. When the standard deviation of simulated noises is between 0.05 and 0.2, the signal-to-noise ratio(SNR) of retrieved turbulence profile with EEMD-SVD denoising is increased by 2.718 7 dB to 6.921 5 dB and the SNR of corresponding EEMD method is increased by 0.168 4 dB to 3.555 4 dB compared with the retrieved profile without denoising. Turbulence profiles retrieved from the undenoised and denoised DCIM lidar measurements and radio-sounding balloons are also compared at two typical time periods. It is found that the maximum SNR of turbulence profiles can separately be increased by 2.526 5 dB and 2.155 6 dB for EEMD-SVD and EEMD method compared with undenoising retrieval profile. The results indicate that the noise reduction effect of EEMD-SVD is superior than EEMD,which it is able to identify and reduce the noises more effectively.The SNR of original signal is greatly improved through EEMD-SVD method, thereby the retrieved atmospheric turbulence profile is achieved more accurately.