Bayesian decompostion of waveform lidar and uncertaitny analysis
The subsquent section is mainly to show how to quantify the uncertaity of waveform processing uinsg Bayesian method. The detailed description of the approach can refer to https://www.researchgate.net/publication/319018940_Bayesian_decomposition_of_full_waveform_LiDAR_data_with_uncertainty_analysis
In the domain of LiDAR applications, the observations or data are inherently subject to various errors such as system setting, system calibration, and range measurement errors. Additionally, the LiDAR vendors often do not clearly state what errors are considered when the data are provided. Thus, uncertainty of “truth” is ubiquitous and inherently present in the realities of LiDAR data modeling.
Furthermore, the models used here are based on the non-linear functions that generally suffer from problem of non-uniqueness, which can generate different parameter combinations given the same observational data and model, or several models can fit observational data at the expense of violating the physical meaning and theoretical assumption of the “real” model.
These problems are more evident for the sophisticated models with multiple peak components in the waveform decomposition. Thus, estimating model uncertainty is imperative for an in-depth understanding of information derived from data and the estimation accuracy.
Overall,the Bayesian decomposition is the intersection of waveform decompostion, Bayesian and non-linear dynamics.
If you find this is interesting, you can go to my github to explore the example code.
Project link: https://github.com/tankwin08/bayesian_decomposition