Two-stage plant identification using spectral signatures and dimensionality reduction
¹Department of Environmental Agricultural Sciences, Faculty of Graduate Studies and Environmental Research, Ain Shams University, Cairo, Egypt. ²Horticulture Department, Faculty of Agriculture, Ain Shams University, 68 Hadayek Shoubra, 11241 Cairo, Egypt. ³National Authority for Remote Sensing and Space Sciences, 23 Joseph Tito Street, El-Nozha El-Gedida, P.O. Box 1564 Alf Maskan, Cairo, Egypt. ?Department of Environmental Management, Institute of Environmental Engineering, RUDN University, 6 Miklukho-Maklaya St, 117198 Moscow, Russia. Corresponding e-mail: heba.mohamed.nagib@iesr.asu.edu.eg
DOI: https://doi.org/10.37855/jah.2025.v27i04.116
Plant species identification using spectral profiles is crucial for agriculture, ecology, and remote sensing, enabling accurate biodiversity monitoring and precision farming. A two-step scheme is proposed in this study for classifying plant species based on spectral reflectance profiles. In the first stage, Principal Component Analysis (PCA) was employed for dimension reduction, and in the second stage, feature selection was achieved using the Least Absolute Shrinkage and Selection Operator (LASSO). Four machine-learning algorithms, namely the Random Forest (RF), Support Vector Machines (SVM), AdaBoost, and k-Nearest Neighbours (KNN), were employed to study the spectral signatures of 31 plant species from three families: Moraceae (with 16 species), Fabaceae (with 8 species), and Euphorbiaceae (with 7 species). Results indicated that the two-stage classification framework enhanced model performance, with SVM and RF attaining accuracies of 89.0% and 87.0%, respectively, compared with 75.0% for KNN and 68.0% for AdaBoost. These results confirm the potential of a two-stage process approach for scalable and high-throughput species identification in ecological and agricultural applications.
