Alternative 2D and 3D Form Characterization Approaches to the Automated Identification of Biological Species

Abstract

Few have sought to compare the performance of alternative types of morphological data for biological species identification. This investigation contrasts results of form characterization via form factors, superposed landmark coordinates, landmark-registered semilandmark outlines, 3D semilandmark networks, and raw digital images for a test set of seven Recent planktonic foraminifer species. While all data types performed better than the qualitative assessment of morphological variation by human taxonomists, landmark-registered semilandmark outlines and raw digital images delivered the best performance in the context of approaches that could reasonably serve as the basis for fully automated species identification systems.