3D-scanning fluorescence imaging of living tissue is in demand for less phototoxic acquisition process. For the imaging of biological surfaces, adaptive and sparse scanning schemes have been proven to efficiently reduce the light dose by concentrating acquisitions around the surface. In this paper, we focus on optimizing the scanning scheme at a constant photon budget, when the problem is to estimate the position of a biological surface whose intensity profile is modeled as a Gaussian shape. We propose an approach based on the Cramér-Rao bound to optimize the positions and number of scanning points, assuming signal-dependant Gaussian noise. We show that, in the case of regular sampling, the optimization problem can be reduced to a few parameters, allowing us to define quasi-optimal acquisition strategies, first when no prior knowledge of the surface location is available and then when the user has a prior on this location.