Medicinal Plants Authenticity Evaluation using an Intelligent Hyperspectral Imaging System Coupled with Biologically Inspired Unsupervised Algorithms: Nepeta Crispa Willd Case Study

Kiani, Sajad; Ayyari, Mahdi

doi:10.22059/ijhst.2025.384543.952

Medicinal Plants Authenticity Evaluation using an Intelligent Hyperspectral Imaging System Coupled with Biologically Inspired Unsupervised Algorithms: Nepeta Crispa Willd Case Study

Document Type : Research paper

Authors

Sajad Kiani ¹

Mahdi Ayyari ²

¹ Biosystems Engineering Department, Sari Agricultural Sciences and Natural Resources University, Sari, Iran

² Horticultural Science Department, Tarbiat Modares University, Tehran, Iran

10.22059/ijhst.2025.384543.952

Abstract

This study evaluated the efficiency of a handheld Hyperspectral Imaging (HSI) system coupled with biologically inspired unsupervised algorithms as a screening technique for the authenticity evaluation of naturally-grown Nepeta Crispa Willd (N.Crispa) samples from on-farm cultivated samples. The volatile oils of 25 samples were isolated with hydrodistillation, and then the Gas Chromatography (GC) analyses were done to determine the total Volatile Organic Compounds (VOCs) as the reference data. On the other hand, the samples' reflectance spectra were captured using the HSI camera and pre-processed using the Savitzky-Golay (SG) algorithm. Principal Component Analysis (PCA) was then applied for the visual discrimination of the samples and the data reduction. Next, two unsupervised algorithms, crisp clustering by the self-organizing Map (SOM) and an automatic clustering based on the artificial bee colony (ABC), were applied to perform the real clustering of the samples. The SOM unified distance matrices explained the changes in spectral characteristics between the N.Crispa samples and indicated the variation of the sample’s VOCs following the GC results. The automatic clustering by the ABC algorithm illustrated its capability to cluster the two main sample groups according to the sample’s spectra. The HSI system combined with the ABC algorithm will provide a novel nondestructive and rapid technique for evaluating the authenticity of naturally-grown Nepeta Crispa Willd samples.

Keywords

Artificial Neural Networks

Clustering

Geographical Origins

Medicinal Plants