ABSTRACT

Full-waveform decomposition is crucial for obtaining accurate satellite-ground distance, the accuracy of which is severely affected by noises. However, the traditional filters all depend on filtering parameters. This paper presents a new and adaptive method for denoising based on empirical mode decomposition (EMD) and Hurst analysis (EMD-Hurst). The noisy full-waveforms are first decomposed into their intrinsic mode functions (IMFs), and the Hurst exponent of each IMF is established by the detrended fluctuation analysis. The IMF is regarded as the high-frequency noise and is deleted if its Hurst exponent is ≤0.5. Both simulated and real full-waveforms were conducted to validate and evaluate the method by comparing with six other IMF selection methods via metrics like waveform decomposition consistency ratio (CR), average error of decomposition parameters, and ICESat/GLAS waveform-parameter product GLAH05. The comparisons show that: (1) under different SNR conditions, EMD-Hurst performs robustly and obtains a higher CR than other EMD based methods; (2) obtains the highest average CR and a relatively lower average error for the echo parameters; and (3) peak numbers and fitting accuracy for GLAH01 are more reasonable and precise than those of GLAH05, which could offer a good reference for the processing on future space-borne full-waveform data.

摘要

全波形分解对于获得精确的卫星到地面的距离至关重要，而卫星到地面的精度会受到噪声的严重影响。但是，传统的过滤器都依赖于过滤参数。本文提出了一种基于经验模态分解（EMD）和赫斯特分析（EMD-Hurst）的自适应去噪新方法。首先将有噪声的全波形分解为其固有模式函数（IMF），然后通过去趋势波动分析建立每个IMF的赫斯特指数。IMF被视为高频噪声，如果其Hurst指数≤0.5，则将其删除。通过与其他六种IMF选择方法（通过诸如波形分解一致性比（CR）的度量）进行比较，进行了仿真和实际全波形验证和评估该方法。分解参数的平均误差和ICESat / GLAS波形参数乘积GLAH05。比较表明：（1）在不同的SNR条件下，EMD-Hurst的性能稳定，并且比其他基于EMD的方法获得更高的CR；（2）获得回波参数的最高平均CR和相对较低的平均误差；（3）GLAH01的峰数和拟合精度比GLAH05更为合理和精确，可为以后的星载全波形数据的处理提供很好的参考。