Because laser diode ceilometer′s backscattering signal is weak and easily disturbed by various noises at the same time, the most important task and difficult point for the signal detection of laser diode ceilometer is how to take effective measures to remove the noise in the backscattering signal. In view of the redundancy reduction capability of the independent component analysis (ICA), fast ICA is proposed to eliminate noise of laser diode ceilometer′s return signals. Since fast ICA requires multi-channel signals, the continuous multiple groups of laser diode ceilometer return signals are used to construct the multi-channel signals, and then the blind source separation (BSS) of fast ICA is applied to the signals. Thus, the virtual sources are extracted one by one, and the noise embedded in the observed signal is removed. The experimental results demonstrate that the method has good effect on removing the noise from laser diode ceilometer′s return signal. Such a fast ICA algorithm has the practical value in processing laser diode ceilometer′s return signals.