Structured light sensor with telecentric stereo camera pair for measurements through vacuum windows

Abstract

Within the Collaborative Research Centre 1153 Tailored Forming a process chain is being developed to manufacture hybrid high performance components made from different materials. The optical geometry characterization of red-hot workpieces directly after the forming process yields diverse advantages, e.g., the documentation of workpiece distortion effects during cooling or the rejection of deficient components in an early manufacturing state. Challenges arise due to the high components temperature directly after forming (approximately 1000 • C): The applied structured light method is based on the triangulation principle, which requires homogeneous measurement conditions and a rectilinear expansion of light. This essential precondition is violated when measuring hot objects, as the heat input into the surrounding air leads to an inhomogeneous refractive index field. The authors identified low pressure environments as a promising approach to reduce the magnitude and expansion of the heat induced optical inhomogeneity. To this end, a vacuum chamber has been developed at the Institute of Measurement and Automatic Control. One drawback of a measurement chamber is, that the geometry characterization has to be conducted through a chamber window. The sensors light path is therefore again affected -in this case by the window's discrete increase of refractive index, and also due to the different air density states at sensor location (density at ambient pressure conditions) and measurement object location (density at low pressure conditions). Unlike the heat induced deflection effect, the light path manipulation by the window and the manipulated air density state in the chamber are non-dynamic and constant over time. The reconstruction of 3-D geometry points based on a structured light sensor measurement directly depends on the mathematical model of detection and illumination unit. The calibration routine yields the necessary sensor model parameters. The window light refraction complicates this calibration procedure, as the standard pinhole camera model used for entocentric lenses does not comprise enough degrees of freedom to adequately parametrize the pixel-dependent light ray shift induced by thick vacuum windows. Telecentric lenses only map parallel light onto a sensor, therefore the window induced ray shift is constant for all sensor pixels and can be directly reproduced by the so-called affine camera model. In this paper, we present an experimental calibration method, and corresponding calibration data and measurement results for a structured light sensor with and without measurement window. The sensor comprises a telecentric stereo camera pair and an entocentric projector. The calibration of the telecentric cameras is conducted according to the well-known affine camera model. The projector is used as feature generator to solve the correspondence problem between the two cameras. The calibration data illustrates that the window refraction effect is fully reproduced by the affine camera model, allowing a precise geometry...