Virtual Inverse fringe projection
Carson Rooms 3 and 4, enter at Carson 3
Andreas Posch - Representing Institute of Measurement and Automatic Control
Industrial quality management requires the inspection of semi-finished and finished goods with respect to their geometric shape. Fringe projection (FP) has been established as a prevalent areal measuring technique, which is economic, flexible and non-contacting. Conventional fringe projection works by projecting structured light patterns – mostly fringes with a specific spatial carrier frequency - onto the optical cooperative surface of a three dimensional workpiece. When these patterns are recorded by a digital camera that resides on a different angle of view than the projector, distorted patterns can be observed which allow for the reconstruction of the object's three dimensional shape by means of the triangulation principle. In order to gain sufficiently reliable and unambiguous data to rebuild the object geometry, several distinct fringe patterns with different spatial frequencies and phases need to be projected and analyzed.
For inverse fringe projection (iFP) an inverse fringe pattern is projected onto the object's surface via a digital projector such that undistorted straight lines emerge in the camera image. This projected inverse fringe pattern must, however, be determined a priori from information about the workpiece's geometry. The use of inverse fringe pattern allows for a much more efficient evaluation of geometry deviations of the workpiece with respect to its expected ideal geometry. With iFP there is no need to project and evaluate more than one distinct pattern which, in contrast to conventional FP, allows for real-time inspection.
Up to now the inverse fringe projection pattern was generated by canceling out fringe distortions in the camera image using a reference object in the measuring setup. This is a very time consuming operation that will block the measurement hardware and requires the physical existence of an appropriate reference object. The creation of a reference object is not only time and finance consuming, but is impossible for prototypes and piece productions.
To overcome the downsides of inverse pattern generation our approach is to computationally obtain the inverse fringe pattern out of the workpiece's CAD data. We intend to perform a full calibration of the real inverse fringe projection system, which only needs to be performed once after system installation, and feed the parameters into a virtual fringe projection system. This allows us to obtain the inverse fringe pattern without any need for a physical reference object nor physical presence of the projection system's hardware. Performing raytracing simulations of the inverse light path (virtually exchanging camera and projector) we expect to be able to directly compute the inverse fringe pattern.
If the object's real shape does not match the CAD model used for inverse pattern computation the straight lines would be observed with some distortions in the camera image. Our aim is to further develop the automated interpretation of inverse fringe projection to quantify the deviation of the workpiece's real shape with respect to the ideal shape of the CAD model in a metric sense.
