It has been demonstrated that decomposing the return waveforms into a mixture of Gaussian components was suitable to achieve the information of the target. But the laser impulse is slightly asymmetric, and approximating the waveforms by using a sum of Gaussians may not be an accurate representation depending on the application and target. This paper focus on the improvement of raw signal modeling and re.ne peak detection greatly increased the number of detected targets as well as their positional accuracy. More complex models than the Gaussian model, such as the Lognormal or generalized Gaussian functions, were introduced and. tted with backscatter waveform of canopy to improve the raw signal by LM algorithm. The results show, through introduction of new echo parameters, it is easy to extract the additional information on the target shape, and calculate reflectivity and roughness.