An Introduction to MuSES and Scene Simulation for AI/ML Applications

On-Demand Webinar

Training robust machine learning algorithms for defense applications often requires vast amounts of accurately labeled real-world data. However, this data can be expensive and time-consuming to obtain, especially for non-visible wavebands. Synthetic target simulation offers a powerful alternative for generating large and diverse image datasets for training.

In this session, we explored how MuSES (Multi-Service Electro-Optic Signature) provides unparalleled accuracy in simulating ground vehicles, manned and autonomous fixed-wing and rotary aircraft, watercraft, human personnel, and other high-fidelity targets in realistic operational environments. Specifically, we walked through its capabilities for realistic environmental modeling, detailed thermal-electrical modeling, human modeling, signature management, and automatic target detection and recognition.

We also demonstrated how to integrate AI/ML techniques with synthetic target simulation to create large training datasets with the desired level of detail, resolution, and variation. This approach draws from an extensive database of 3D target models with accurate geometric features, optical surface properties, and internal heat sources.

Key Takeaways

• Realistic Environmental Modeling: Simulate battlefield conditions. This includes solar loading, atmospheric effects, and terrain interaction.
• Detailed Thermal-Electrical Modeling: Incorporate electronic heat sources and coupled battery and photovoltaic power systems. Phase change materials and other sophisticated target details are also included.
• Human Modeling: Include thermoregulating humans with realistic poses and clothing in scenes.
• Signature Management: Explore how optical surface properties and thermal management systems affect target detectability. This is especially useful in cluttered scenes.
• Automatic Target Detection/Recognition: Investigate how effectively AI/ML algorithms detect and recognize embedded targets. In addition, see how key variables impact performance.

Who Should Watch

• Defense Engineers and Researchers working on ISTAR (intelligence, surveillance, target acquisition, and reconnaissance) efforts.
• Data Scientists and Image Analysts supporting vehicle system development and evaluation.
• Program Managers focused on soldier vulnerability and mission planning. This also includes search and rescue operations and automated security applications.

Presenters

Eli Datema

Thermal/CFD Engineer

Eli Datema is a Thermal/CFD Engineer at ThermoAnalytics, Inc., where he develops thermal and EO/IR models for defense and commercial applications using TAITherm, MuSES, and CoTherm. Based in the Calumet office, Eli specializes in creating high-fidelity scene simulations that support a range of use cases, including AI/ML algorithm development. He began his career at ThermoAnalytics as an intern in 2020 and joined full-time in 2021 after earning his B.S. in Mechanical Engineering from Michigan Technological University.

Jacob Hendrickson

Thermal/CFD Engineer

Jacob Hendrickson is a Thermal/CFD Engineer at ThermoAnalytics, Inc., where he works within the Engineering Services group based in Novi, Michigan. He focuses on thermal model preparation and validation, including full vehicle thermal models developed using TAITherm. Hendrickson also supports the company’s defense-side work, running EOIR simulations for target assets and their associated synthetic scene-scapes using MuSES. His human thermal modeling contributions include developing a new manikin model test bed design and creating a novel infant physiology model for TAITherm’s Human Modeling Extension. He holds a B.S. in Mechanical Engineering from Michigan Technological University and an M.B.A.