相干光时域反射仪(COTDR)是测量海底光缆工作情况的设备，由于相干瑞利噪声和偏振噪声的影响，实际测量得到的COTDR检测曲线中往往含有比较剧烈的振幅抖动，使得自动识别光缆上发生的反射事件变得非常困难。文章提出了一种有效的基于小波变换和线性相关系数的滑动平均算法，降低了COTDR曲线上的振幅抖动，与传统的小波去噪算法相比，有效地提高了测量曲线的平滑度，且成功实现了事件的自动检测。

The Coherent Optical Time Domain Reflectometer (COTDR) is an efficient tool to measure the state of submarine optical cable. However, it is difficult to detect the events on optical cable automatically due to the amplitude fluctuation on COTDR curve, which is caused by coherent Rayleigh noise and polarization noise. An efficient COTDR signal processing algorithm based on wavelet transform, linear dependent coefficients and moving averaging method is proposed. The amplitude fluctuation can be reduced and the curves obtained by the proposed algorithm is much smoother than traditional wavelet denoise algorithm. It is shown that almost all the events can be found in experiments.