Implementation of a LIDAR signal processing algorithm for high resolution measurements of wind fields

Abstract

High spatial resolution measures of wind fields are an important topic in LIDAR research in view of the impact they can have in different areas. A method for generating a spectrogram from the analysis of a LIDAR signal has been set up with the use of a new parsimonious description. Coupled with a minimisation algorithm, the method improves the spatial resolution of the radial velocity.

This new method is compared with a validation method using the covariance matrix of the simulated signal. This validation shows that a continuous method has a better accuracy than a discrete generation due to the approximations caused by the discretisation of the frequencies of the signal.

Using this new generation, an optimisation algorithm is used in order to fit those frequencies more precisely. The goal of this algorithm is to minimise a chi-square criterion based on the experimental spectrogram and the spectrogram generated at each iteration of the optimisation. The initial guess is given by a Gaussian fit of the experimental spectrogram. This algorithm greatly improves the spatial resolution, in a relatively short time. Possible improvements on the generation, the initialisation and the optimisation are also discussed in this thesis.