Beyond traditional hyperspectral imaging: exploring spectral DNA encoding and spectral velocity for enhanced food quality detection.

Abstract

This study examines the use of hyperspectral imaging for the identification of stale food items by analyzing minute changes in their spectral signatures. An algorithm is proposed for the detection of subtle alterations in spectral signatures and is validated through intra-class classification comparisons among various stages of adulterating food samples acquired using a spectroradiometer. The analysis reveals that the spectral angle mapper proves effective for inter-class classification of consumable food items but faces challenges in classifying slight changes in spectral signatures within the same category. In contrast, DNA encoding demonstrates reliability, despite the generated code-words being independent of the actual intensity of received reflectance at each band. DNA encoding can provide insights into the nature of absorbance or reflectance at each band, making it a valuable tool for intra-class classification. Additionally, a novel concept called spectral velocity is introduced for subclass pattern matching. This method of single-pixel analysis relies on artificially constructed nD-vectors derived from spectral signatures. The findings suggest that the combination of hyperspectral imaging and DNA encoding offers a valuable tool for the quality assurance of consumable food items and demonstrates its potential for ensuring food safety and quality, ultimately contributing to human health.