Uncertainty-aware 3D Reconsturction for Synthetic Environment

Abstract

Proposed Title: Uncertainty-aware 3D photogrammetric reconstruction for synthetic environmentAbstract: Synthetic Environments (SE) aim to replicate the real-world scenario to a degree that simulated activities can accurately predict the potential engagement in the field, such that decisions made in SE can be adapted in real-world practices. Typically, field-level geo-specific models are generated through image-based photogrammetric approaches, where errors propagated through typical photogrammetric processes may avoidably impair the geometric fidelity the models, while these errors remain uncharacterized in synthetic environment. In addition, to create realistic simulations, SE needs to know detailed object categories such as light poles, traffic signs, and bare terrains, to model the respective effects for immersive experiences. These applications require the high-resolution 3D geo-specific data to be accuracy-wise characterizable, semantically resolvable, and photo-realistically renderable when being incorporated into SE simulators. This project aims to develop a set of fundamental techniques to improve the error characterization, semantic resolution, and realism ofgeo-specific 3D photogrammetric models for synthetic environment, to introduce an uncertainty-aware geo-specific 3D asset creation paradigm for enhance the fidelity of virtual training and rehearsal.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 24, 2023
Source ID
N000142312670

Entities

People

  • Rongjun Qin

Organizations

  • Office of Naval Research
  • Ohio State University
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Computer Vision.
  • Military Training and Readiness Simulation