Point cloud based robot cell calibration

This paper presents a method to automatically calibrate an articulated robot arm using measured point cloud data. The method captures the inner structure of complex engineering objects from measured datasets. In the developed workflow first the point cloud is segmented, and then the CAD models of the objects in the work cell are recognized and fitted onto the segmented point cloud. To boost the computational efficiency of the method, parallelization was performed by applying general-purpose programming of the graphics processing unit. In this paper the calibration of a UR5 robot is carried out using measured data of a Kinect sensor.