INTERGEO Agenda - english

The objective of a UAS survey using cameras and/or laser scanners is often the three-dimensional reconstruction of the captured scene. One typical output is a dense 3D point cloud. This point cloud is frequently converted into a mesh, and the reconstructed surfaces are textured using high-resolution image data. The resulting 3D meshes consist of polygonal elements—typically triangles—that define an explicit geometric surface. As such, meshes provide a metrically accurate and explicitly defined geometry that can be readily used as input for CAD systems, graphics engines, and standard 3D modeling software. As an alternative or complementary representation, 3D Gaussian Splats have recently gained significant attention. In this approach, a scene is represented by millions of small, overlapping three-dimensional ellipsoids (Gaussian splats) that store information about color, transparency (opacity), size, and orientation. This representation enables highly photorealistic rendering, accurately reproducing fine details, reflections, transparent objects, and thin structures that are often difficult to represent with conventional 3D meshes. In contrast to meshes, however, Gaussian splats do not provide explicit geometric surfaces and are therefore only of limited suitability for engineering applications that require precise geometric models or physics-based simulations. The presentation provides an overview of the current state of the art in the acquisition and generation of both representation types. It further discusses their respective strengths and limitations and highlights potential application areas, with a particular focus on geodetic and geomatics applications.