Volumetric Sensor Alignment
Multi-sensor spatial alignment is typically achieved via marker-based methods, where all views align with respect to some markers and then registered among themselves. Traditionally this involved moving a checkerboard pattern within the capturing space, or in the case of retro-reflective markers, a wand. We move beyond such tedious, error-prone processes and instead employ a structure-based calibration without any markers. Users simply need to assemble a structure out of commonly found packaging boxes and place it within the capturing area for all sensors to see. For that we rely on data-driven correspondence establishment for the inital matching [1, 2], and global optimization for estimating a solution with respect to the coordinate system of the structure.
The calibration software needs to be installed after setting up VolCap, with the necessary infomation available in the StructureNet Installation section. Then, the structure needs to assembled, which is an one-time process as it can be easily re-used. The assembly process is described at the Structure Assembly section. Finally, the calibration process itself is explained in the Multi-sensor Calibration section.
[2] Sterzentsenko, V., Doumanoglou, A., Thermos, S., Zioulis, N., Zarpalas, D. and Daras, P., 2020, March. Deep soft procrustes for markerless volumetric sensor alignment. In 2020 IEEE Conference on Virtual Reality and 3D User Interfaces (VR) (pp. 818-827). IEEE.