A new approach for the effective processing of hyperspectral images: application to pushbroom-based anomaly detection and compression systems

Abstract

In this Thesis work we have dealt with the issue around the onboard execution of multiple hyperspectral image analysis techniques onto the same piece of hardware, paving the way for the real-time performance of the hyperspectral image processing. The main objective of this Thesis is to provide the research community with a set of common core operations that extract useful information from the HSIs for many applications; such as anomaly detection, target detection, lossy compression, classification, and unmixing. On the one hand, it results in many benefits in view of hardware acceleration in terms of a reduction in the execution times, hardware resources and above all, in human endeavours. Concerning this latter, it implies the studio and analysis of only a single mathematical approach, which consequently permits to focus the efforts from a methodological and productivity points of view. On the other hand, it also permits the simultaneous execution of many different tasks at the same time with the advantage of sharing the most computationally intensive operations. As a consequence, it promotes the decrease in the amount of computational resources compared with those scenarios in which different state-of-the-art algorithms are independently executed for each targeted processing analysis.