Optical remote sensing of air pollutant: an analysis using wavelet representation

A multiresolution analysis using wavelet representation is applied to the identification and quantification of air pollutants from remotely sensed data. The wavelet representation provides a means to decompose an absorption spectrum with both scale and wavenumber localization. Wavenumber localization is essential in the analysis of absorption spectra because the presence or absence of absorption at certain wavenumbers are unique for each pollutant. If the analyzing wavelets are orthonormal, the wavelet representations between two successive resolutions are orthogonal. Using the wavelet representation, the analysis for air pollutant identification is carried out first on a relatively low resolution followed progressively by analyses at higher resolutions. The coarse-to-fine resolution analyses are performed recursively until the identification process converges. Good performance in terms of processing time and accuracy has been achieved using this approach. Sample analysis of some measured spectra are included