ICESat-2 is the successor to NASA’s ICESat (Ice, Cloud and land Elevation Satellite) mission scheduled to be launched in 2018. The new photon counting LiDAR onboard ICESat-2 introduced new challenges to the estimation of biomass and its dynamics, especially for the abundant photon noise in the atmosphere and below the ground. In order to remove the ambient noise and get a better detection of the canopy and the ground, this paper establishes an approach to identify potential signal from ambient noise automatically. The framework is based on the classic geodesic active contours method. Previous studies have suggested that this technique is very sensitive to initial contour, so we adopted NASA’s surface-finding algorithm to get the expected initialization for contour evolution. Observations from MABEL (Multiple- Altimeter Beam Experiment LiDAR), which is the ICESat-2’s high altitude airborne demonstrator, were used to validate this approach. The results showed that the potential signal photons were about 22% among the whole photons compared with about 78% background noise even in the night flight situation. The signal-to-noise ratio is expected to be smaller in the daytime flight situations, making it more difficult to distinguish the canopy. The results demonstrated that this technique can identify the potential signal photons effectively with error rate less than 4.2%. The proposed approach is appropriate for the present airborne simulated data with a high accuracy for flat surface with dense canopy. Future work will be focused on the stability and general applicability of this approach over large areas and different ground surfaces.