Target Identification and Detection Using Hyperpectral Signature Transformation

Hyperspectral imaging has emerged as a powerful technique for target detection and material identification across various fields, from remote sensing to environmental monitoring. This paper proposes a spectral signature matching method for detecting a known target (i.e., sea water) in hyperspectral images. By analyzing pixel-by-pixel spectral characteristics, spectral matching algorithm computes the similarity between the spectral signature of each pixel and that of known target materials.The proposed method employs the Jasper Ridge hyperspectral data from an urban landscape to demonstrate the algorithm’s effectiveness, with results revealing high accuracy in target detection even in heterogeneous environments where various land cover types (like water, vegetation, urban structures, soil, and rocks) are intermixed within the same area. This highlights the algorithm’s robustness in discerning similar spectral features. The experimental results demonstrate that the Jeffries-Matusita Spectral Angle Mapper (JMSAM) outperformed the Spectral Angle Mapper (SAM) and the Normalized Spectral Similarity Score (NS3) algorithms, by achieving the lowest MSE of 0.056933.