Extrinsic Calibration of UAV and Multi RGB-D Cameras with Non-overlapping by using an External Camera

This paper presents a novel extrinsic calibration approach for rigid RGB-D cameras with non-overlapping view and UAV, which enables the tracking mission and autonomous flight to support the pilots during the control operation. The checkerboards with QR-marker are used as calibration patterns. Furthermore, a QR code attached on the top of UAV is used to define the UAV reference coordinate system. By importing an additional camera, the overlapping with the two rigid cameras and UAV is created, a 3D structure of calibration patterns is reconstructed from the sequence of images from all cameras by using bundle adjustment algorithm. Simultaneously, the pose of each camera at different frame optimized by minimizing the corresponding 3D projection errors. Furthermore, according the 3D coordinate of the QR-code corners in the reconstructed map, the transformation from UAV reference coordinate system to 3D structure can be obtained. Thus, the extrinsic parameters, i.e. relative position and orientation between fixed cameras and a predefined UAV coordinate system can be determined. The experiments have shown that the presented calibration algorithm provides reliable and accurate results. Meanwhile, the redundancy, reliability, and robustness of the proposed method for different setups of the rigid cameras are evaluated.