Palynology is widely used in both academic and industrial research for correlation and interpretation of subsurface geology on both local and regional scales. Although most datagathering tools for the subsurface have undergone major technological developments during recent decades, palynological research has remained virtually unchanged. With increasing demand for faster and more detailed palynological analyses, palynology is at risk of being left behind. The answer could lie in adopting strategies of digitalization and Artificial
Intelligence (AI) originally developed for medical research. The technology of scanning microscope slides into digital high-resolution images has already been developed, and AI software specifically developed for palynology can be utilized to examine the digital images and detect, identify and quantify the fossil content. Identification of the 3-dimensional microfossils in a 2-dimenzional view will require several morphological parameters, some of which may not be present in every scanned specimen located. The image recognition software thus needs to be able to identify fossils from different angles, preservation levels and fragmentation stages, along with partially obscured or folded specimens. Morphological variations within taxa, evolution and sub-division also needs to be taken into considerationby the AI. This project explores the advantages and disadvantages of digitally scanned palynological slides and the use of AI software recognition. It will establish if the digital scanned slides have the resolution needed to be used for quantitative analysis and any limitations that hinder taxonomic assignments compared to transmitted light microscopy. In
addition, the project will design a preparation protocol in order to produce the most reliable slides for digital scanning. We will attempt to design and develop an AI software for clustering and classification of the microfossils based on deep-learning based algorithms fordetection and segmentation of palynomorphs, and algorithms based on autoencoders for extracting features predictive of different fossil taxa.