3D surface reconstruction of entomological specimens from uniform multi-view image datasets

摘要

Modeling of 3D objects and scenes have become a common tool in different applied fields from simulation-based design in high-end engineering applications (aviation, civil structures, engine components, etc.) to entertainment (computer-based animation, video-game development, etc.). In Biology and related fields, 3D object modeling and reconstruction provide valuable tools to support the visualization, comparison and even morphometric analysis in both academical and applied tasks. Such computational tools, usually implemented as web-based virtual reality applications, significantly reduce the manipulation of fragile samples, preventing their damage and, even, their complete loss. On the other hand, they allow to take the morphological properties of physical specimens to the digital domain, giving support to common entomology tasks such as characterization, morphological taxonomy and teaching. This paper addresses the problem of producing reliable 3D point clouds from the surface of entomological specimens, based on a proved approach for multi view 3D reconstruction from high resolution pictures. Given the traditional issues of macro-photography for small sized objects (i.e. short depth of field, presence of subtle and complex structures, etc.), a pre-processing protocol, based on focus stacking, supported the generation of enhanced views obtained by an acquisition device specifically designed for this work. The proposed approach has been tested on a sample of six representative subjects from the Entomological Collection of the Centro de Biosistemas, Universidad Jorge Tadeo Lozano (Colombia). The resulting point clouds exhibit an overall good visual quality for the body structure the selected specimens, while file sizes are portable enough to support web based visualization.