ObjSplat: Geometry-Aware Gaussian Surfels for Active Object Reconstruction

Autonomous high-fidelity object reconstruction is fundamental for creating digital assets and bridging the simulation-to-reality gap in robotics. We present ObjSplat, an active reconstruction framework that leverages Gaussian surfels as a unified representation to progressively reconstruct unknown objects with both photorealistic appearance and accurate geometry. Addressing the limitations of conventional opacity or depth-based cues, we introduce a geometry-aware viewpoint evaluation pipeline that explicitly models back-face visibility and occlusion-aware multi-view covisibility, reliably identifying under-reconstructed regions even on geometrically complex objects. Furthermore, to overcome the limitations of greedy planning strategies, ObjSplat employs a next-best-path (NBP) planner that performs multi-step lookahead on a dynamically constructed spatial graph. By jointly optimizing information gain and movement cost, this planner generates globally efficient trajectories. Extensive experiments in simulation and on real-world cultural artifacts demonstrate that ObjSplat produces physically consistent models within minutes, achieving superior reconstruction fidelity and surface completeness while significantly reducing scan time and path length compared to state-of-the-art approaches. Note to Practitioners—This paper addresses the challenge of autonomous, high-fidelity digitization of physical objects, a capability essential for digital cultural heritage preservation and XR asset creation. Currently, existing automated systems often rely on pre-programmed trajectories that cannot adapt to unknown shapes or utilize local planning strategies that result in inefficient, redundant movements. We present ObjSplat, a unified active reconstruction system that overcomes the inefficiencies of manual scanning and the limitations of existing predefined trajectories or greedy automation methods. By leveraging Gaussian surfels and a geometry-aware evaluation pipeline, the system reliably identifies under-reconstructed regions on complex objects (e.g., hollow or thin structures). Unlike traditional view-by-view strategies, our multi-step Next-Best-Path (NBP) planner optimizes global movement, significantly reducing operation time and redundant motion. The framework is ready for deployment on a robotic arm equipped with RGB-D sensors to produce physically consistent, watertight models within minutes, with future potential to address complex optical properties and multi-robot collaboration.