Light Detection and Ranging (LiDAR) is a form of remote sensing that utilizes laser scanners to produce a 3D point cloud of an environment by recording the number of laser pulse returns and measuring the backscattered energy as a function of time. LiDAR transect data were collected over the Monterey Peninsula and the Point Lobos Reserve. An experiment was conducted in the creation of a transect, a very high point density profile, by restricting the scan mirror with the initial goal of better understanding foliage penetration by LiDAR. Because of the high point density of the transect, the data were binned to create synthetic waveforms and to help reduce redundant points. However, the binning introduces sharpness in the data that distorts the typical wave shape in the synthetic transforms. A Bayesian Markov random field model captures the structure in the dataset and helps to offset the sharpness introduced during the binning. After fitting a Markov random field model using Markov chain Monte Carlo, classification methods were applied to distinguish objects in the landscape. These techniques should extend to true waveform data.
|