A gravitational model for plant classification using adaxial epidermis texture

Abstract

The leaves are very informative plant organs. They are extensively used in plant anatomical studies focusing taxonomy. Their both inner and outer structures provide very discriminant features from vegetal species. In this study, we propose using images from adaxial epidermis for plant classification. The adaxial epidermis is a very variable region in a plant leaf cross-section. It differs in color, number of layers and presence/absence of hypodermis. To accomplish this task, we propose combining complexity analysis methods with a gravitational collapsing system to extract texture features from adaxial epidermis samples. Experimental results show that this combination of techniques surpasses traditional and state-of-the-art methods in both grayscale and color images of adaxial epidermis.