Utilizing machine-learning-based 3D image analysis for classifying charred grape seeds to the varietal level

Abstract

Abstract Grapevine (Vitis vinifera L.) is an essential part of the oldest group of fruit trees around which horticulture evolved, currently includes thousands of cultivars, grown at numerous climatic conditions. Discrimination between these varieties has been traditionally conducted using ampelography, and in recent decades mostly by genetic analysis. However, when aiming to identify archaeobotanical remains, which are mostly charred- with extremely low genomic preservation, the application of the genomic approach is rarely successful. As a result, variety-level identification of most grape remains is currently prevented. Because grape pips are highly polymorphic, several attempts were made to utilize their morphological diversity as a classification tool, mostly using 2D image analysis technics, aiming to utilize these methods for the identification of fresh and archaeological specimens. Here, we present for the first time a highly accurate varietal classification tool, using an innovative and accessible approach for 3D seed scanning. The suggested classification methodology is machine-learning-based, using a complete set of 3D data obtained for each seed, applied with the Iterative Closest Point (ICP) registration algorithm and the Linear Discriminant Analysis (LDA) technique. This methodology achieved classification results of ca. 90-99% accuracy when trained by fresh seeds to test unknown fresh seeds. Moreover, the classification of charred seeds reached up to 100% accuracy when trained by charred seeds. Based on this approach, our long-term aim is to develop a computerized classification tool for the identification of grape and possibly other species and varieties. Such a tool can significantly improve the fields of archaeobotany, as well as general taxonomy.