Evaluating Interior Orientation Estimability in Multimedia Photogrammetry - A practical guide on object-invariant bundles

Abstract. Bundle adjustment in multimedia environments, such as underwater or through refractive interfaces, poses unique challenges for parameter estimation due to increased correlations between interior orientation and refractive parameters. This contribution investigates the estimability and correlation of these parameters in object-invariant multimedia bundles by presenting both a simulated and a real-world dataset. Using a strict ray tracing bundle adjustment approach, we analyze how water depth, surface tilt, and parameter set selection influence correlations and numerical stability. Statistical metrics - including correlation matrices, parameter significance tests, and variance inflation factors (VIF) - are evaluated for their effectiveness in diagnosing problematic configurations. Results show that while traditional metrics like σ₀ may not reveal instability, VIF and correlation analysis provide practical additional procedures for identifying robust parameter estimations. The findings offer a workflow for practitioners, highlighting optimal parameter configurations and the limitations of statistical diagnostics in multimedia photogrammetry.