Quality Assessment for Hyperpspectral Airborne Systems

Abstract

In order to produce validated, and therefore reliable, products based on hyperspectral airborne imagery, it is necessary to first characterise the data quality itself. Since airborne hyperspectral imagery is usually acquired in several partially overlapping flight-lines, one can use this information redundancy to retrieve the imagery’s internal variability. However, in practice, the geo-locations assigned to each pixel is not error-free. A first step therefore consists in retrieving the relative variability of the pixel’s geo-locations, hence providing pairs of pixels imaging the same geo-locations. In a second time, these pairs of pixels are used to obtain ’quality indicators’ assessing the quality of land-products in terms of reproducibility. Several other cated indicators are also proposed, as well as means to obtain their corresponding confidence intervals and their suitability is discussed. The third stage of this analysis consists in studying whether it is possible to use these reliability results to improve the mapping results. A generic approach based on optimising reliability over the mapping parameters is proposed. The final step consists of using specific flight-lines setups in order to retrieve the reflectance' variability by means of a blind-deconvolution approach.