Investigation of interpretation possibilities of spectral high-dimensional measurements by means of principal component analysis: a concept for physical interpretation of those measurements

Publisher Website
Google Scholar
Abstract
Subject of the paper is the investigation of the information content of high dimensional multispectral remote-sensing measurements in the VIS-NIR region for ocean-atmosphere problems. The final goal of such measurements is the separation of atmospheric influence and the retrieval of detailed information of water constituents. Primary questions appearing during interpretation process are: how many independent parameters can be found from the measurements? what is the physical sense of these parameters? what is the accuracy of the parameters? One has to take into account that the properties of the measuring device, like channel position, bandwidth, number of channels and measurement accuracy have a great influence on the interpretation. One possible method to get answers to the above questions is the Principle Component Analysis (PCA). A problem in PCA is the physical interpretation of the mathematically obtained results - Eigenvalue, Eigenvector and Principle Components. Because the results of PCA interpretation depend on the statistical properties of the measurement data, they must be mapped back to the absolute measurement quantities (radiances). To get a physical interpretation of the PCA results a detailed investigation with a simulated data set using a simplified (but nonlinear) model was realized (atmosphere after Gordon, Sturm, water reflectances after Sathyendranath, Morel, Prieur). It will be presented a concept, how in-situ measurements can be involved into interpretation model with PCA.© (1993) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

This publication has 0 references indexed in Scilit: