Influence of the structured illumination frequency content on the correspondence assignment precision in stereophotogrammetry

Stereophotogrammetric 3D shape measurement using structured illumination is an established class of methods for industrial inspection. One essential step in the measurement process for all stereophotogrammetric techniques is the assignment of corresponding points between the stereo views. As the purpose of the used structured illumination is to ease and improve the correspondence assignment, the choice of said sequence is of utmost importance. The precision of the correspondence assignment directly affects the noise of the final point cloud and therefore this assignment should be conducted with the highest precision possible. Depending on the chosen structured illumination sequence, different degrees of freedom for the pattern design exist and may affect the precision of the correspondence assignment and thus the noise of the 3d point cloud. In our contribution we want to discuss the influence of the frequency content of the structured illumination for a scheme employing bandlimited statistical patterns, which have been fruitfully used for highspeed applications in the past years. To evaluate the limits of the correspondence assignment accuracy we created a simple numerical signal-detector model. Using this model the correspondence assignment in dependence of the chosen structured illumination can be compared to ground truth-data. Furthermore, the noise of point clouds in real measurements is investigated to validate the results of the used simulation. Therefore, illumination sequences using different spatial frequency bands are created and projected onto a reference object. Afterwards, the noise of the resulting pointcloud is evaluated. The results indicate that it is advisable to optimize the pattern design depending on the used sensor and object properties.