Efficient representation of 3D spatial data for defense-related applications

TL;DR

This paper compares traditional and modern neural methods for representing 3D geospatial data in defense, highlighting their strengths and weaknesses, and proposes a hybrid system combining both approaches for improved performance.

Contribution

It provides a comprehensive analysis of 3D data representation techniques and introduces a hybrid architecture that leverages the advantages of both traditional and neural methods.

Findings

Traditional models offer robust geometric accuracy.

Modern neural methods produce high-fidelity visuals.

A hybrid approach combines geometric integrity with visual detail.

Abstract

Geospatial sensor data is essential for modern defense and security, offering indispensable 3D information for situational awareness. This data, gathered from sources like lidar sensors and optical cameras, allows for the creation of detailed models of operational environments. In this paper, we provide a comparative analysis of traditional representation methods, such as point clouds, voxel grids, and triangle meshes, alongside modern neural and implicit techniques like Neural Radiance Fields (NeRFs) and 3D Gaussian Splatting (3DGS). Our evaluation reveals a fundamental trade-off: traditional models offer robust geometric accuracy ideal for functional tasks like line-of-sight analysis and physics simulations, while modern methods excel at producing high-fidelity, photorealistic visuals but often lack geometric reliability. Based on these findings, we conclude that a hybrid approach is the most promising path forward. We propose a system architecture that combines a traditional mesh scaffold for geometric integrity with a neural representation like 3DGS for visual detail, managed within a hierarchical scene structure to ensure scalability and performance.