Joint Anomaly Detection and Spectral Unmixing for Planetary Hyperspectral Images

Abstract

Abstract-Hyperspectral images are commonly used in the context of planetary exploration, especially for the analysis of the composition of planets. As several instruments have been sent throughout the Solar System, a huge quantity of data is getting available for the research community. Among classical problems in the analysis of hyperspectral images, a crucial one is unsupervised non-linear spectral unmixing, which aims at estimating the spectral signatures of elementary materials and determining their relative contribution at a sub-pixel level. While the unmixing problem is well studied for earth observation, some of the traditional problems encountered with earth images are somehow magnified in planetary exploration. Among them, large image sizes, strong non-linearities in the mixing (often different from those found in the earth images) and presence of anomalies are usually impairing the unmixing algorithms. This paper presents a new method that scales favorably with the problem posed by this analysis. It performs an unsupervised unmixing jointly with anomaly detection capacities, and has a global linear complexity. Non linearities are handled by decomposing the hyperspectral data on an overcomplete set of spectra, combined with a specific sparse projection, which guarantees the interpretability of the analysis. A theoretical study is proposed on synthetic datasets, and results are presented over the challenging 4-Vesta asteroid dataset.

Can Terrestrial Restoration Methodologies be Transferred to Planetary Hyperspectral Imagery? A Quantitative Intercomparison and Discussion

“…Most subsequent analysis techniques will suffer from degraded performance in the presence of noises. For instance, endmembers will be estimated incorrectly when unmixing, thereby leading to nonsensical abundances [22]. Albeit planetary HSIs are contaminated seriously, but their noise types are still similar to those of terrestrial images.…”

Can UAVs fill the gap between in situ surveys and satellites for habitat mapping?

“…The unmixing algorithm used to estimate abundances was SAGA+ (Nakhostin et al, 2016), which is based on the geometric concept of finding the simplex that embeds data. The simplex is calculated in a feature space associated with a kernel, which is useful when analyzing non-linear mixtures.…”

Bilinear normal mixing model for spectral unmixing

“…For example, it can provide abundance information to improve the performance of classification [6, 7], and provide endmember information for target detection [8]. The number of SU applications is enormous, including research topics such as urban environment [9, 10], natural environment [11, 12], agriculture [13] and ecosystem [14], planetary exploration [15, 16] and so on.…”