Pattern sequences for fast absolute phase retrieval with application in the handheld operation of structured light sensors

Areal 3D scanners based on structured light projection are widely used for inspection tasks in manufacturing processes. For accurate measurements of object geometries, common structured light sensors project a sequence of periodic patterns onto the surface. By subsequent evaluation of these patterns, the local phase of the patterns is reconstructed in a phase retrieval step. The local phase encodes the geometry information and allows for the reconstruction of 3D point clouds in combination with a calibration of the sensor. However, motion of an object during the projection of the pattern sequence leads to large deviations in the reconstructed local phase and thus the 3D point cloud. For dynamic measurements of moving objects, several single-shot techniques have been proposed. While these techniques only require a single pattern for 3D measurements, phase reconstruction suffers when measuring non-diffuse reflective surfaces. In order to evaluate the feasibility of different approaches, sequences of periodic patterns with different lengths are compared based on static and dynamic measurements of a contour standard with a non-diffuse surface using a fiber-endoscopic structured light sensor. Based on the evaluations, pattern sequences are parametrized for dynamic measurements using the flexible sensor head of the endoscopic system. A proof of concept of simultaneous localization and mapping for the handheld operation of the endoscopic sensor head is demonstrated, which can be used for the navigation in narrow spaces in the future