On the Geometric Accuracy of Implicit and Primitive-based Representations Derived from View Rendering Constraints

2025-09-12

AI Summary

This study addresses the stringent geometric accuracy requirements for 3D object reconstruction in space robotics applications. We systematically compare the geometric fidelity of implicit (K-Planes) and explicit (Gaussian Splatting, Convex Splatting) novel-view synthesis methods and investigate the impact of appearance embedding on geometric quality. Experiments on the SPEED+ space-scene dataset reveal three key findings: (1) Appearance embedding significantly improves rendering quality but yields no substantial gain in geometric accuracy; (2) Convex Splatting achieves superior geometric fidelity with more compact primitives and stronger robustness to clutter—making it better suited than Gaussian Splatting for safety-critical tasks; (3) Embedding integration in explicit methods reduces the required number of primitives, enhancing geometric representation efficiency. To our knowledge, this work is the first to identify the appearance–geometry decoupling phenomenon and empirically validate the structural advantages of Convex Splatting for space-based 3D reconstruction.

Technology Category

• Computer Vision: 3D Computer Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

We present the first systematic comparison of implicit and explicit Novel View Synthesis methods for space-based 3D object reconstruction, evaluating the role of appearance embeddings. While embeddings improve photometric fidelity by modeling lighting variation, we show they do not translate into meaningful gains in geometric accuracy - a critical requirement for space robotics applications. Using the SPEED+ dataset, we compare K-Planes, Gaussian Splatting, and Convex Splatting, and demonstrate that embeddings primarily reduce the number of primitives needed for explicit methods rather than enhancing geometric fidelity. Moreover, convex splatting achieves more compact and clutter-free representations than Gaussian splatting, offering advantages for safety-critical applications such as interaction and collision avoidance. Our findings clarify the limits of appearance embeddings for geometry-centric tasks and highlight trade-offs between reconstruction quality and representation efficiency in space scenarios.

Problem

• Evaluating geometric accuracy of implicit vs explicit view synthesis methods
• Assessing role of appearance embeddings for 3D reconstruction fidelity
• Comparing representation efficiency for space robotics applications

Innovation

• Compares implicit and explicit view synthesis methods
• Evaluates appearance embeddings' impact on geometric accuracy
• Demonstrates convex splatting advantages for space applications