The Projected Pattern Correlation (PROPAC) technique is an optical 3D measurement technique, which is able to measure deformations of diffuse scattering surfaces spatially. The achievable measurement accuracy can be less than one micrometer. PROPAC enables us to determine surfaces of objects in absolute coordinates as a difference to a defined reference surface.
The measurement principle is shown in the adjoining sketch: a stationary projection unit projects a random dot pattern under a small angle onto the surface. A stationary camera records the dot pattern under a triangulation angle. Due to the random characteristic of the dot pattern, its cross-correlation is always unique. This allows us to calculate the local cross-correlation between the dot pattern on the deformed surface and on the reference surface and, therefore, the local pattern shift on the surface. For known orientation of the projection unit and the camera, it is possible to reconstruct the orthogonal surface deformation by means of the measured pattern shift. In doing so, it is possible to use the well-developed and robust cross-correlation algorithms originally developed for Particle Image Velocimetry (PIV) applications. Due to the geometrical measurement setup, it is necessary to apply the Scheimpflug condition to the camera and the projection system respectively.
Using high-power light sources enables us to use an exposure time of a few microseconds for image acquisition. This makes real-time measurements of surfaces in production lines possible. Compared to other conventional optical full-field measurement techniques, PROPAC is considerably fast, because it only needs one single measurement image for the calculation of the deformation.
The measurement principle of using cross-correlation on random dot patterns is applied to specular surfaces by means of the Reflected Pattern Correlation (REPAC). In this case, the stationary camera records an image of an also stationary screen with a random dot pattern indirectly, using the specular surface as a mirror. This principle is shown in the adjoining sketch.
The measured pattern shift in case of the REPAC technique can be introduced by a variation of height or a variation of declination. Therefore, the surface has to be observed with two cameras from different directions. Using an appropriate reconstruction algorithm, information about height and declination for every measured point can be obtained. This can be a big advantage if one wants to calculate the curvature of the surface (e.g. for eyeglasses) because in this case only one derivation is needed.