3D Reconstruction of Reindeer Antlers Using a Low-Cost Optical Camera system and Gaussian Splatting

Abstract. Caribou and reindeer antlers provide valuable insights for biodiversity, environmental monitoring, and regenerative medicine. However, traditional methods for measuring antlers rely on manual, time-intensive techniques prone to human error and ethical concerns due to physical animal interaction. This project aims to address these limitations by developing an automated, non-invasive 3D reconstruction pipeline utilizing optical imagery and Gaussian Splatting (GS). The system employs a multi-camera setup to capture high-resolution optical imagery, addressing challenges such as motion artifacts and inconsistent lighting. Prior approaches using time-of-flight cameras suffer from lower quality data, impact from subject movement and interference between cameras. GS-based methods using optical imagery are able to handle these conditions more effectively. Initial development and testing in a laboratory environment replicating the animal pen has shown strong results. With comparison to a reference laser scanner derived point cloud, a cloud-to-cloud mean distance of 3.4-3.7 mm was achieved. The system has been installed at the animal pen and one season of collection was performed; preliminary results showing good quantitative accuracies. The project’s objectives include the production of high-quality 3D reconstructions of reindeer antlers, with the capability to handle subject movement. Expected outcomes include an automated 3D modeling pipeline, high-quality antler datasets spanning 2 growth cycles, providing valuable data for interdisciplinary research. These results will progress the understanding of antler growth dynamics and provide a scalable solution for future endeavors. By integrating optical imaging techniques with GS-processing, this research offers a novel approach to generating 3D models of antlers with wide-ranging applications.