Propuesta de un método computacional para la detección de asbesto en imágenes hiperespectrales a partir de la similitud diferencial espectral

Abstract

Considering that one of the challenges of hyperspectral imaging is identifying methods that enable the effective and efficient detection of materials, this article proposes a new method for detecting asbestos in hyperspectral images based on spectral differential similarity. This method determines how closely the spectral signature of a given pixel matches the spectral signature of asbestos. The proposed method was implemented using open-source libraries such as spectral, numpy, pandas, and matplotlib. Compared to the correlation method, it detected 0.813% fewer vegetation pixels. In terms of computational efficiency, the proposed method was 4.27 times faster than the correlation method. The results indicate that the proposed method demonstrates adequate efficacy and excellent efficiency, making it a strong candidate for integration into tools for processing and analyzing hyperspectral images in academic and industrial domains.