S5-4 - Data-efficient optical 4D measurements of moving objects with stereo event cameras and tailored structured illumination
- Event
- 23. ITG/GMA-Fachtagung Sensoren und Messsysteme 2026
2026-06-09 - 2026-06-10
Nürnberg - Band
- Vorträge
- Chapter
- Event-Based Vision
- Author(s)
- C. Freitag, A. T. Khusainova, S. Heist, P. Kühmstedt, G. Notni - Fraunhofer IOF, Jena, A. Stark, C. Franke - Friedrich-Schiller-Universität Jena, Jena
- Pages
- 105 - 112
- DOI
- 10.5162/sensoren2026/S5-4
- ISBN
- 978-3-910600-11-9
- Price
- free
Abstract
We present an optical 4D (3D+time) measurement system for data-efficient reconstruction of shape and motion of rigid objects. In contrast to frame-based active stereo systems, which repeatedly resample surfaces at high rates, our approach exploits event cameras to break the tight dependence of scanning rate to object speed. The system combines a stereo pair of event cameras with a projector that generates a periodically rotating contrast edge to sparsely probe the measurement volume. In contrast to frame-based cameras, event cameras only report brightness changes and thus adapt their measurements to the dynamics of the scene. The moving contrast edge sparsely probes the measurement volume. Triangulation of temporally corresponding events in both cameras yields 3D surface points with individual timestamps. We specifically target sub-scan dynamics, i.e. significant object motion during a single slow scan of the contrast edge. Allowing significant sub-scan motion fundamentally breaks the assumptions of frame-based structured-light systems and introduces two key challenges: ambiguous temporal correspondences and motion-induced 3D rolling-shutter distortions. Our reconstruction framework addresses both effects. First, a geometric 4D model of the structured illumination is estimated and used to resolve ambiguity. Second, we estimate the object’s motion and warp all 4D points to a common reference time, compensating the rolling-shutter effect. Experiments with a translating rigid object demonstrate that, despite low scan rates and significant sub-scan motion, the proposed method recovers the object’s shape with sub-millimetre deviation from a static reference.