A computer vision methodology for pollen classification using SEM: a case study with medicinal plant species

Pollen grains of plant species have unique morphological characteristics. The characteristics, such as variability in pollen shape, size, and microscopic surface features, can be effectively used in the identification of plant species. This approach can be helpful in the identification of medicinal plant species that have a rich biodiversity. This study presents a computer vision-driven framework designed for the segmentation and classification of pollen images acquired through scanning electron microscopy (SEM). This framework has broad applications in palynological fields, including taxonomy, paleoecology, and biodiversity monitoring. While the methodology is universally applicable, we illustrate its effectiveness using a representative dataset consisting of SEM images from 28 medicinal plant species, selected for their morphological diversity and availability. The dataset features 269 images dedicated to segmentation and 5842 images for classification across 28 distinct classes, which have been compiled into an open-access resource titled MPalyn (Medicinal Pollen and Palynology SEM Database). This study demonstrates a broad application in offering a scalable and adaptable tool for automated pollen analysis, which has the potential to advance research in different fields of palynology.