Evaluation of military camouflage using hyperspectral satellite data

Evaluating concealment capabilities objectively and reproducibly has always been a challenge in the camouflage community. The development of a robust assessment method is crucial to evaluate current and future camouflage fabrics. However, evaluating samples directly in the field can be costly and time consuming. That is especially true for remote or inaccessible regions. This study aims to provide a rapid and repeatable preliminary assessment method using hyperspectral satellite data to characterize natural and artificial backgrounds. A spectral matching pipeline is then applied to compare these environmental signatures with lab measurements of the camouflage samples. This approach provides a quick evaluation of concealment capabilities based on spectral similarity, reducing the need for in situ measurements. Typical hyperspectral data used in this work encompasses woodland, transition and urban environments within the same image frame. A pixel per pixel approach of spectral matching using well-known metrics such as Spectral Information Divergence (SID) and Spectral Angle Mapper (SAM) allow to compare spectra in various conditions. The end result is a heatmap of spectral similarities covering the whole frame, allowing to see, within a hyperspectral image, which environments are better or worse for a given camouflage fabric. This work can then be extended to more diverse geographical regions, spanning from the forests and fields of Ukraine to African deserts and tropical jungles, with no need to travel on site. The data covers the 380-2500nm spectral range, giving a peek at multispectral concealment capabilities in the Visible (VIS), Near Infrared (NIR) and Short-Wave Infrared (SWIR) domains. This study takes a purely hyperspectral approach, the focus is therefore put on hyperspectral sensor threats.