A Fast Calibration Implementation for Multiple Depth Cameras and Manipulator Based on Invariance of the Linear Transformation

For the safety in the human-robot interaction and co-existence, perceiving dynamic unknown surroundings around the manipulator is basic requirements. To fully monitor the surroundings around the manipulator, multiple depth (RGB-D) cameras are simultaneously used to address the problem of occluded areas. The fast and efficient estimation of the rigid relationship between the manipulator and multiple cameras coordinate frames is essential. In this study, a method based on invariance of the linear transformation was proposed to realize fast calibration for multiple RGB-D cameras and manipulator. Firstly, the centers of the calibration ball were extracted directly from the depth image. Secondly, the motion straight lines of the ball were, respectively, fitted in the robot and multiple cameras coordinate frames. Finally, the analytical solution was obtained by using the invariance of the linear transformation. Two Kinect cameras and a KUKA IIWA manipulator were employed to conduct experiments, and experimental results were presented to validate the performance of the proposed calibration method for multiple RGB-D cameras and manipulator.