Improved empirical mode decomposition based denoising method for lidar signals

详细信息    查看全文

• 作者：Pengfei Tian ; Xianjie Cao ; Jiening Liang ; Lei Zhang ; ^{zhanglei@lzu.edu.cn" class="auth_mail} ; Nana Yi ; Liying Wang ; Xiaoping Cheng
• 关键词：Lidar ; Signal denoising ; Empirical mode decomposition
• 刊名：Optics Communications
• 出版年：30 August 2014
• 年：2014
• 卷：325
• 期：Complete
• 页码：54-59
• 全文大小：1419 K

文摘

Based on the physical significance of intrinsic mode functions (IMFs) and the noise component removed from the empirical mode decomposition (EMD) method, the denoising process of the lidar (CE370-2, Cimel) backscattering signal is analyzed in detail. Two parameters, typical range (TR) and low-frequency fraction (LFF) are suggested as the reference principles to decide how many high-frequency IMFs should be removed as noise. TR represents the major spatial range of each IMF, which increases with the decrease in the frequency of IMFs; LFF represents the relative value of the low-frequency component of the removed component, which increases as more IMFs are removed. The simulated signals show that the cloud layer altitudes and intensities impact little on the noise reduction processes. Based on an appropriate amount of lidar data, thresholds for TR and LFF are provided, respectively, for various weather conditions: 0.330 and 0.276 for clear sky conditions, 0.460 and 0.517 for cloudy conditions, 0.331 and 0.316 for dusty conditions, and 0.327 and 0.310 for polluted conditions. These thresholds are applied to the automatic data-denoising algorithm. Only 3.9% of the data encounters a numerical calculation error for the clear sky conditions, and the percentage increases to 8.5% for cloudy conditions, which is also acceptable. It turns out that the automatic EMD denoising method has a better denoising performance than that of the wavelet method.