Flying Triangulation - towards the 3D movie camera

TL;DR

This paper discusses advancing Flying Triangulation sensors for 3D movie cameras by increasing data density and measurement depth, addressing fundamental limits and false line indexing issues for real-time, motion-robust 3D data acquisition.

Contribution

It introduces approaches to overcome line indexing ambiguities and explores information-theoretical limits to enhance single-shot 3D sensing capabilities.

Findings

Current sensors project about 10 lines with 1000 pixels each.

Significant potential to increase line density and measurement depth.

Proposed solutions aim to approach information-theoretical limits of 3D sensors.

Abstract

Flying Triangulation sensors enable a free-hand and motion-robust 3D data acquisition of complex shaped objects. The measurement principle is based on a multi-line light-sectioning approach and uses sophisticated algorithms for real-time registration (S. Ettl et al., Appl. Opt. 51 (2012) 281-289). As "single-shot principle", light sectioning enables the option to get surface data from one single camera exposure. But there is a drawback: A pixel-dense measurement is not possible because of fundamental information-theoretical reasons. By "pixel-dense" we understand that each pixel displays individually measured distance information, neither interpolated from its neighbour pixels nor using lateral context information. Hence, for monomodal single-shot principles, the 3D data generated from one 2D raw image display a significantly lower space-bandwidth than the camera permits. This is the price one must pay for motion robustness. Currently, our sensors project about 10 lines (each with 1000 pixels), reaching an considerable lower data efficiency than theoretically possible for a single-shot sensor. Our aim is to push Flying Triangulation to its information-theoretical limits. Therefore, the line density as well as the measurement depth needs to be significantly increased. This causes serious indexing ambiguities. On the road to a single-shot 3D movie camera, we are working on solutions to overcome the problem of false line indexing by utilizing yet unexploited information. We will present several approaches and will discuss profound information-theoretical questions about the information efficiency of 3D sensors.